Ecology, 85(6), 2004, pp. 1708±1724 ᭧ 2004 by the Ecological Society of America

CLIMATIC AND HUMAN INFLUENCES ON FIRE REGIMES OF THE SOUTHERN SAN JUAN , COLORADO, USA

HENRI D. GRISSINO-MAYER,1,4 WILLIAM H. ROMME,2 M. LISA FLOYD,3 AND DAVID D. HANNA3 1Department of Geography, University of Tennessee, Knoxville, Tennessee 37996 USA 2Department of Forest Sciences, Colorado State University, Fort Collins, Colorado 80523 USA 3Environmental Studies Program, Prescott College, Prescott, Arizona 86301 USA

Abstract. Fire severity, frequency, and extent are expected to change dramatically in coming decades in response to changing climatic conditions, superimposed on the adverse cumulative effects of various human-related disturbances on ecosystems during the past 100 years or more. To better gauge these expected changes, knowledge of climatic and human in¯uences on past ®re regimes is essential. We characterized the temporal and spatial properties of ®re regimes in ponderosa pine forests of the southern San Juan Mountains of southwestern Colorado by collecting 175 ®re-scarred tree samples from nine sites across a wide range of topographic settings. All tree rings and ®re scars were dated using standard dendrochronological techniques. Fire-free intervals were statistically modeled using the Weibull distribution to provide quantitative measures that characterized the historical range of variation in pre-EuroAmerican ®re regimes. Fires during our reference period were more frequent in the low ponderosa pine forests (6±10 yr) than in the high elevation, mixed conifer forests (18±28 yr). Fires at lower were predominantly low-severity, isolated ®res. Fires during some years (e.g., 1748) were spatially extensive throughout the entire range. Intervals that delimited signi®cantly long ®re-free periods ranged from 10±19 yr (low elevation) to 27± 50 yr (high elevation). Fire histories were similar between the eastern and western portions of the mountain range, although we found signi®cant evidence of topographic isolation on ®re regimes at one site. Pre-1880 ®res primarily occurred in the dormant season, and we found no temporal changes in past ®re seasonality. We found no compelling evidence that Native Americans in¯uenced ®re regimes in our study sites. We found a hiatus in ®re occurrence between 1750 and 1770 that we believe was likely related to weakened El NinÄo-Southern Oscillation activity, an extended series of cool-phase Paci®c Decadal Oscillation events, and weakened monsoonal moisture, all possibly en- trained in an invasive air mass typical of locations that are more northerly. In addition, pre-1880 ®res occurred during years of severe drought, conditioned by above average moisture conditions in preceding years. The 20th century is characterized by a near complete absence of ®res (®re-free interval of Ͼ100 yr), suggesting future wild®res may be more widespread and ecologically severe compared to pre-1880 ®res. Key words: climate±®re interactions; dendrochronology; disturbances, human; ®re history; ®re regimes; mixed conifer; ponderosa pine; San Juan Mountains, Colorado, USA.

INTRODUCTION of many size and age classes (Cooper 1960, Swetnam Ponderosa pine (Pinus ponderosa Dougl. ex Lawson) and Baisan 1996, Fule et al. 1997). Pure ponderosa is widespread throughout much of western North pine forests in eastern Oregon, Washington, and west- America, extending from the foothills to middle ele- ern Montana had similar ®re regimes (Arno 1980, Agee vations of many mountain ranges and plateaus, and 1993, Everett et al. 1994, Heyerdahl et al. 2001), as did low-elevation ponderosa pine forests near the for- grows within a wide variety of site conditions and cli- est±grassland ecotone in northern Colorado (Veblen et matic regimes (Oliver and Ryker 1990). Recent re- al. 2000) and the Black of South Dakota and search has demonstrated that ®re regimes in ponderosa Wyoming (Brown and Sieg 1996, 1999). Fire-free in- pine forests varied substantially through time and tervals also appear to lengthen with increasing eleva- across space prior to EuroAmerican settlement in the tion within any given region, although many local ex- mid to late 1800s. Most forests of pure ponderosa pine ceptions to this pattern have been documented (Arno in Arizona and New were characterized by fre- 1980, Swetnam and Baisan 1996, Veblen et al. 2000). quent, low-severity ®res that maintained open forests Fire regimes also varied temporally with different patterns in different regions. Many studies have doc- Manuscript received 15 July 2002; revised 2 April 2003; ac- cepted 22 April 2003; ®nal version received 4 September 2003. umented a close relationship between changes in pon- Corresponding Editor: D. Urban. derosa pine ®re regimes and changes in climate. Studies 4 E-mail: [email protected] conducted in Arizona, New Mexico, and Colorado, for 1708 June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1709 example, show a reduction in the occurrence of wide- could be more aligned with that found in Arizona and spread ®res in the late 1700s to early 1800s, perhaps New Mexico. as a result of reduced El NinÄo-Southern Oscillation Third, a sense of urgency exists because the behavior (ENSO) activity (Cleaveland et al. 1992, Swetnam and of future wild®res and reintroduction of ®re by man- Betancourt 1998, Grissino-Mayer and Swetnam 2000, agement agencies require knowledge of past ®re re- Veblen et al. 2000). A similar hiatus in ®re occurrence gimes, including ®re frequency, spatial extent, season- is apparent in more northerly reconstructions of ®re ality, and response by ®re to climatic factors (Swetnam from the Black Hills region during 1706±1753 (Brown et al. 1999). This knowledge provides a template on and Sieg 1996) and at Cheesman Lake in the Front ®re attributes as they likely existed prior to Euro- Range of Colorado during 1723±1851 (Brown et al. American in¯uence, by which investigators can eval- 1999). Because these regions lie outside the known uate current and future ®re behavior. The evaluation is geographic in¯uence of ENSO activity, some other cli- all the more important because increasing global tem- mate mechanism may be responsible for this conspic- peratures and associated changes in precipitation pat- uous, and earlier, gap in ®re occurrence. Other studies terns are occurring, primarily from enhanced levels of have documented effects on ®re regimes associated greenhouse gases caused by human activity. Several with Native American practices in western Montana studies have clearly demonstrated changes in ®re re- (Barrett and Arno 1982), southern New Mexico (Kaye gimes associated with expected changes in future cli- and Swetnam 1999), and southern Arizona (Seklecki mate, including changes in ®re frequency, extent, sea- et al. 1996). Fire frequency also increased in a portion sonality, and timing of the ®re season (Balling et al. of the western Black Hills in the late 1700s, coincident 1992, Wotton and Flannigan 1993, Bergeron and Flan- with the arrival of the Sioux people to the region (Fisher nigan 1995). To gauge the in¯uence of these expected et al. 1987). The relative roles of lightning and native changes in climate on future ®re regimes, investigators peoples on ®re regimes in the uplands of the Southwest must know the past relationship between climate and were recently summarized by Allen (2002). ®re across many spatiotemporal scales. We addressed four primary questions with our re- We reconstructed the temporal and spatial properties construction of past ®res. (1) How did pre-1880 ®res of ®re regimes in ponderosa pine forests of the southern vary temporally within the SJM? We were speci®cally San Juan Mountains (SJM) in southwestern Colorado interested in analyzing ®re±climate and ®re±human re- for several reasons. First, a large and conspicuous gap lationships, their variability over time, and whether a exists in southwestern Colorado where no studies have hiatus in ®res occurred in the late 1700s to early 1800s. reconstructed past ®res, despite a broad network of ®re (2) How did pre-1880 ®res vary spatially within the history sites throughout western (Hey- SJM? We expected to ®nd longer intervals at higher erdahl et al. 1995, Swetnam and Baisan 1996). The elevations, in sites having topographic barriers to ®re locations of our study sites are important for bridging spread, and in the generally wetter eastern portion of ®re regimes in the Southwest with those found in lo- the SJM. (3) Were pre-1880 ®re-free intervals in the cations that are more northerly. This contiguous net- SJM similar to the shorter intervals that characterized work will facilitate interpretations of possible climatic forests in Arizona and New Mexico or to the longer forcing mechanisms that operated in the past to affect intervals found in northern Colorado? (4) To what ex- ®re regimes across broad spatial scales. tent did Native American people in¯uence pre-1880 A second goal was to determine whether ®re regimes ®re regimes? Native Americans may have augmented in the SJM were aligned to those found in more north- the natural ®re frequency and altered the seasonality erly or southerly locations. The SJM exist near the of past ®res with intentionally set ®res. northern periphery of the semiarid Southwest. The SJM therefore may experience a shorter ®re season than STUDY AREA more southerly locations, perhaps contributing to fewer The San Juan National Forest covers 757 000 ha in ®res throughout the year. Furthermore, the fuel com- the southern and western portions of the massive San plex in southwestern Colorado consists of forest with Juan Mountain Range in southwestern Colorado, USA understory dominated by shrubs; this is more similar (Fig. 1). Elevations range from 1800 m to Ͼ4200 m, to the ponderosa pine forests in northern Colorado than and a wide variety of geologic substrates and topog- to the forests with grass-dominated understory of raphy are present. Ponderosa pine and mixed conifer northern Arizona and New Mexico. Thus, it is possible forests cover extensive areas at the lower to middle that ®re regimes in southwestern Colorado share pat- elevations (2100±2900 m) of the San Juan Mountains, terns of ®re frequency with northern areas. The con- and exhibit a complex mixture of species compositions, tinental location, however, would contribute to higher disturbance regimes, and local histories (DeVelice et summer temperatures, greater convective activity, a al. 1986, Romme et al. 1992). Ponderosa pine is the higher frequency and greater intensity of thunder- major canopy species in all of the forests within this storms, and perhaps cause more lightning ignitions. vegetational zone, but associated species vary with el- Thus, it is also likely that the ®re regime of the SJM evation and aspect (Floyd-Hanna et al. 1996). At lower 1710 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6

FIG. 1. The San Juan National Forest of southwestern Colorado, showing site locations. See Table 1 for codes and full site names. elevations and on xeric aspects, ponderosa pine forms rioration was already reported (DuBois 1903, Dishman pure stands in the overstory. Six of our sites were lo- 1982, Dahms and Geils 1997:32). This early heavy cated in such stands (Table 1). At higher elevations and grazing probably reduced biomass and changed com- on mesic aspects, the understory is dominated by munity composition in the herbaceous strata of the for- shrubs, especially Gambel oak (Quercus gambelii ests (Arnold 1950, Madany and West 1983, Fleischner Nutt.). Additionally, we sampled three stands in mixed 1994). Logging of ponderosa pine forests began in the conifer forests (Table 1), representing the warm, dry, late 1800s, accelerated during the 1920s, and by the mixed conifer forests growing at lower elevations ad- mid-20th century, nearly every accessible stand had jacent to pure stands of ponderosa pine. The cool, wet, been affected (Dishman 1982, Smith 1982; J. S. Red- mixed conifer forests found at higher elevations and ders, personal communication). Most logging empha- lacking ponderosa pine were not considered here. sized selective removal of the large, high quality trees. Prior to EuroAmerican settlement, the major inhab- itants of the San Juan Mountain region were the Ute METHODS people (Callaway et al. 1986, Ellis 1996). Little is Site selection known about the prehistory of the Utes, and it is not certain when they moved into the region (Duke 1995). We selected nine study sites representing the range Available evidence suggests that they were relatively of elevation, topography, and habitat types within the few in number and were scattered over an enormous ponderosa pine and warm, dry, mixed conifer zones of area (Bolton 1950). Oral histories indicate that the Utes the San Juan Mountains (Fig. 1, Table 1). These sites set ®res on occasion (M. Colyer, personal communi- ensured that a comprehensive east-west gradient was cation), but almost nothing is documented about spe- represented. Candidate sites were ®rst inspected to en- ci®c Ute ®re practices. We therefore know little about sure that each had visible evidence of ®re (i.e., trees the kinds or magnitudes of impacts that the Utes may with multiple scars). Site SMI is located near the end have had on local ecosystems during the period of in- of a long narrow ridge, with steep barren slopes on the digenous settlement. west and south that potentially serve as barriers to ®re The ®rst permanent EuroAmerican settlers in the San spread. None of the other sites has any topographic Juan Mountains were miners who arrived in the 1860s features that would likely limit ®re spread into or out (Smith 1996), followed by cattlemen in the 1870s, and of the site. Site TRS is adjacent to an archaeological sheep herders soon after (Dishman 1982). Livestock site representing a Ute camping area of the 18th and grazing was unregulated and apparently heavy in many 19th century (Duke 1995). Although Native American areas. As early as the 1890s, widespread range dete- people traveled throughout the San Juan Mountain area, June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1711

TABLE 1. Characteristics of sites where ®re history was reconstructed in the San Juan Mountains, Colorado, USA, arranged from highest elevation to lowest elevation.

No. tree-ring samples³ Study site Area Dead or and code Location Elevation (m) (ha) Topography Vegetation² Live remnant Total Burnette Canyon 37Њ35Ј N, 2740±2840 50 narrow ridge and MC with PIPO 5 3 8 (BCN) 108Њ08Ј W adjacent slopes Taylor Creek 37Њ36Ј N, 2530±2650 100 steep slopes and MC with PIPO 9 7 16 (TCK) 108Њ13Ј W deep ravines Monument 37Њ19Ј N, 2470±2560 200 moderate slopes MC with PIPO 3 15 18 (MNT) 107Њ12Ј W and ravines Smoothing Iron 37Њ32Ј N, 2470±2560 200 broad isolated PIPO with QUGA 9 14 23 (SMI) 108Њ21Ј W ridge top Benson Creek 37Њ08Ј N, 2440±2560 200 moderate slopes PIPO with QUGA 9 14 23 (BCK) 106Њ52Ј W and ravines Hermosa Creek 37Њ28Ј N, 2410±2560 50 steep slopes and PIPO/QUGA with 21 2 23 (HCK) 107Њ50Ј W deep ravines some PSME Turkey Springs 37Њ18Ј N, 2440±2470 200 gentle slopes, pure PIPO 4 24 28 (TRS) 107Њ10Ј W nearly ¯at Plateau Creek 37Њ38Ј N, 2410±2440 150 gentle slopes, pure PIPO 1 20 21 (PLT) 108Њ28Ј W nearly ¯at Five Pine Canyon 37Њ41Ј N, 2250±2375 200 gentle slopes and PIPO/QUGA with 51015 (FPC) 108Њ41Ј W shallow ravines PJ ²MCϭ mixed conifer; PSME ϭ Pseudotsuga menziesii (Douglas-®r); PIPO ϭ Pinus ponderosa (ponderosa pine); QUGA ϭ Quercus gambelii (Gambel oak); PJ ϭ pinyon juniper. ³ Total number of tree-ring samples analyzed from all sites ϭ 175; these include 66 from living trees and 109 from dead trees or remnants of trees. we are unaware of similar intensive use in the vicinity pression can be caused by physical damage to foliage of our other study sites. and to roots, a growth release can be caused by ®re- induced nutrient enhancement and/or reduction of Field methods neighboring competitors, while resin duct formation is At each site, we collected between 10 and 30 sam- a common response by conifers to physical trauma (i.e., ples, targeting logs, snags (standing dead trees), rem- resinosis; Brown et al. 1992, Brown and Swetnam nants (portions of eroded logs), stumps, and living trees 1994, Wimmer 2002). that displayed visible sequences of multiple ®re scars Trees with multiple scars occasionally have discon- on the basal portion of the tree bole. We used a chain tinuous tree-ring records due to repeated burning and saw to extract complete cross sections from the bole removal of the outermost tree rings and perhaps any through the ®re-scarred surface. Because ®re-scarred embedded ®re scars. This removal would affect cal- living trees provide information on 20th century ®re culations of percentage of trees scarred when assessing regimes, small partial sections were cut from selected how widespread any particular ®re was within any site. living trees (Arno and Sneck 1977, Baisan and Swet- To ensure valid results from the statistical analyses that nam 1990). We also recorded information about each assessed ®re extent, we documented whether tree rings tree (e.g., diameter at breast height, crown condition, were ``recorder years'' or ``non-recorder years'' (Gris- and signs of injury or damage) that would aid the tree- sino-Mayer 1999). ``Recorder years'' are those tree ring dating process. Areas sampled ranged from 50 ha rings found on intact, noneroded sections that formed to 200 ha. after the initial scarring event, and contain a ®re scar or could have contained one had a ®re occurred. ``Non- Laboratory methods recorder years'' are those tree rings that preceded the All cross sections were reassembled and mounted on initial scarring event, or were too eroded, decayed, or plyboard if necessary, then sanded using progressively burned to provide ®re history information. All statis- ®ner sandpaper to produce high quality, polished trans- tical analyses were conducted over segments of re- verse surfaces on which the cellular structure of the corder years only. tree rings could be readily identi®ed under standard 7± 10ϫ magni®cation. All tree rings were crossdated, us- Fire seasonality ing reference chronologies previously developed for The dominant season of past ®re occurrence was the northern Rio Grande area and Mesa Verde National evaluated by recording the intraannual position of the Park (Dean and Robinson 1978). In addition to the year ®re scar within the tree ring (Dieterich and Swetnam of scar formation, we also recorded possible ®re-in- 1984, Baisan and Swetnam 1990). If the position of duced zones of growth releases or growth suppressions the scar could not be determined with accuracy, it was and whether resin ducts were present. Growth sup- not included in the analysis. We assigned calendar 1712 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6 months to these intraannual positions based on previous er 1999), which in all instances indicated superior ®ts studies that investigated the cambial phenology of of the Weibull distribution compared to the assumed Southwestern tree species growing near Mesa Verde normal distribution. We assessed central tendency in National Park (Fritts et al. 1965), and from a pilot study the distribution of ®re-free intervals using two mea- initiated in 1995 at our Hermosa Creek site (W. H. sures. The Weibull Median Interval (MEI) is the in- Romme, unpublished data): terval associated with the 50th percentile of the dis- tribution, while the Weibull Modal Interval (MOI) is 1) EE: a scar located in the ®rst third portion of the the interval associated with the maximum area under earlywood (mid-June to early July). the probability density function. The range of historical 2) ME: a scar located in the second third portion of variability was delimited by calculating the Lower and the earlywood (early July to mid-July). Upper Exceedance Intervals (LEI and UEI; see Gris- 3) LE: a scar located in the latter third portion of the sino-Mayer 1999). Variance in ®re-free intervals was earlywood (mid-July to late July). assessed using the coef®cient of variation. 4) L: a scar located in the latewood (late July to To assess the ability of ®re to spread within each August). study site, we additionally computed statistics for ®res A dormant season ®re scar (D) positioned between that scarred at least 25% of the recorder trees within tree rings could have occurred either in early spring the area and compared these to statistics based on all (April to mid-June) or late summer/fall (August and ®res, regardless of the number of trees that recorded September); winter snowfall makes it unlikely ®res oc- ®res. Percentage scarred classes have been used exten- curred during other months. Because no physiological sively in previous studies to indicate ®res that were evidence allowed us to determine in which of these two perhaps more widespread within individual sites (Swet- seasons the ®re occurred, a dormant season ®re scar nam 1990, Swetnam and Baisan 1996, Fule and Cov- was interpreted as a spring ®re. Assigning a calendar ington 1997, Veblen et al. 2000). year to a dormant season scar, however, was occasion- To assess temporal changes in ®re regimes, we di- ally assisted by observing the ®re scar positions for the vided the period of analysis for the entire composited same year on other nearby samples. For example, if San Juan data set into two independent periods based some samples from the same site had a few scars in on apparent changes in ®re-free intervals as seen in the the latewood portion of the tree ring, then any dormant composite graphs. The difference in mean intervals be- season scars on other samples from the same site could tween periods was statistically evaluated using a Stu- be assigned to the late summer/fall season for the same dent's t test on interval data that were ®rst normal trans- year. formed to reduce effects of skewness, and then con- verted to standard normal distributions. In addition, Examining ®re-free intervals post-1880 ®re regimes were compared to pre-1880 ®re For each sample, we recorded ®re years, scar posi- regimes using critical threshold descriptive statistics tion, any injuries or anomalous ring features, inner/ (the UEI and maximum interval) to assess the degree outer ring dates, and whether the rings were recorder of hazard that currently exists in the forests of the San or non-recorder years. This information was then en- Juan Mountains (Grissino-Mayer 1999). tered into an FHX2 database (Grissino-Mayer 2001). For each site, we created graphs that depicted the tem- Evaluating the ®re±climate association poral and spatial patterns of past ®res (Dieterich 1980). To identify climate conditions that acted as precur- The beginning year for statistical analyses at each site sors to ®re occurrence, we used superposed epoch anal- was that year when at least two samples recorded a ®re ysis (Swetnam and Betancourt 1992, Swetnam and (which ranged from 1654 to 1748; Swetnam and Baisan Baisan 1996), which evaluates climate conditions prior 1996, Grissino-Mayer 1999). The year 1880 was cho- to and during ®re years. Average climate conditions sen as the ending year because human-related distur- are calculated for each year prior to (up to t Ϫ 5), during bances that altered ®re regimes became pervasive be- (t ϭ 0), and after (to t ϩ 2) the year of ®re. Con®dence ginning in the 1880s throughout much of the American intervals for assessing statistical signi®cance were cal- West. culated using bootstrapping techniques on events ran- A key objective in the analysis of ®re-free intervals domly drawn from the population of observations. is to assess how frequently ®res occurred historically. Because instrumental weather records are too brief We composited (i.e., ®re dates from all samples com- to assess pre-1880 ®re regimes, we used a limber pine posited to a single time series) then modeled ®re-free tree-ring chronology as a proxy for precipitation. These interval data using the Weibull distribution, which has trees were collected from a site in the southern portion been shown to be an effective model of Southwestern of the San Juan Mountains (R. Adams, L. Baxter, and ®re regimes (Swetnam and Baisan 1996, Grissino-May- J. Fairchild, unpublished data). Correlation analysis er 1999). Goodness-of-®t of the Weibull distribution to between this chronology and precipitation for Durango, the ®re-free interval data was measured using the Kol- Colorado (1900±1990, available from the Western Re- mogorov-Smirnov goodness-of-®t test (Grissino-May- gional Climate Center, Reno, Nevada) revealed these June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1713

FIG. 2. Master ®re chronologies for all nine study sites in the San Juan Mountains from 1650 to 1994, arranged from highest elevation (BCN) to lowest elevation (FPC). Horizontal lines represent the time span for individual trees; solid lines indicate recorder years; dashed lines indicate non-recorder years (see Methods: Laboratory methods). Short vertical bars depict dated ®re scars.

trees respond predominantly to February±July total RESULTS rainfall (r ϭ 0.62, P Ͻ 0.0001), with June precipitation Temporal patterns in past ®re having the greatest single-month relationship (r ϭ 0.53, P Ͻ 0.0001). Therefore, tree-ring indices used in the Fires were relatively frequent between 1680 and superposed epoch analysis can be used to infer spring/ 1880 for ponderosa pine forests in the San Juan Moun- summer rainfall conditions (i.e., values above/below tains as a whole (Fig. 2). Fire frequency dropped dra- 1.0 represent wetter/drier spring/summer conditions). matically after 1880 (at all sites with suf®cient numbers 1714 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6 of samples), coinciding with the onset of heavy grazing, sistent with the wide range in its measures of central although one site (HCK) recorded a widespread ®re in tendency (MOI ϭ 3 yr while MEI ϭ 13 yr). 1890. The paucity of ®res during the 20th century is Few of our sites showed unambiguous evidence of remarkable. Six sites (BCN, TCK, and MNT at high a hiatus in ®re occurrence in the late 1700s to early elevation, SMI and BCK at mid elevation, and TRS at 1800s (Fig. 2) as documented at numerous other sites low elevation) had no widespread ®re after 1880, a ®re- in the Southwest (Swetnam 1990, Swetnam and Baisan free interval of at least 114 years (1881±1994). The 1996, Grissino-Mayer and Swetnam 2000). Instead, we HCK site had an unprecedented (within its 300-yr re- observed a clear hiatus from about 1750 to about 1770 cord) ®re-free interval of 100 years (1890±1989). At in the composite graphs for all sites (the horizontal the two remaining sites (PLT and FPC), sample sizes lines in Fig. 4A), especially when ®ltered to show the were too low during the 20th century to assess ®re widespread ®res (Fig. 4B). To determine whether a frequency. change in ®re frequency had occurred after the hiatus, Several key patterns emerged among the measures we divided the entire reference period into two shorter of central tendency. The median ®re-free interval was, sub periods, 1680±1752 and 1767±1880. The (normal without exception, either equal to or shorter than the transformed) mean ®re-free interval for the 1680±1752 mean ®re-free interval because the median observation period was signi®cantly longer than the mean interval in a data set is generally more resistant to ``outlier for the 1767±1880 period for all ®res (P ϭ 0.04) and drag.'' In addition, the two central measures based on for the more widespread ®res (P ϭ 0.03; Table 3). the Weibull distribution were generally shorter than the Spatial patterns in past ®re mean ®re-free interval, with the Weibull Modal Interval (MOI) generally shorter than the Weibull Median In- A clear elevational gradient in ®re occurrence is ap- terval (MEI). There was little difference between the parent (Fig. 2). As expected, ®res were not as common median ®re-free interval and the Weibull Median In- in the higher elevation mixed conifer stands (BCN and terval except for the higher elevation sites, which were TCK) as they were in drier, pure ponderosa pine stands at lower elevations (PLT and FPC), while mid-elevation characterized by longer intervals. At the SMI site, the sites with mixed pine and Douglas-®r (HCK and SMI) MOI was much lower than the MEI, re¯ecting the ef- had intermediate values (Table 2). Many ®res were fects of a higher number of shorter intervals (Fig. 3). widespread within individual sites (i.e., they scarred a These four measures clearly formed a regular pattern, large percentage of sampled recorder trees for that with few exceptions: mean ®re-free interval Ͼ median year), but were likely low-severity ®res for so many ®re-free interval Ն Weibull Median Interval Ͼ Weibull trees to have survived. Two of the pure pine sites were Modal Interval. Of these four measures, the Weibull anomalous, however. Statistics for the low-elevation Median Interval appeared less in¯uenced by clusters TRS site are more like those of the higher elevation, of short or long intervals in the distributions, and we ponderosa pine±Douglas-®r sites, while the BCK site, therefore recommend this as a preferred measure of at a relatively high elevation, has ®re-free interval sta- central tendency in ®re-free interval analyses. tistics similar to lower elevation sites. Similar patterns emerged among the statistics that In the lower and middle elevation sites, MEI values describe the range in ®re-free intervals among the sites. for ®res that scarred Ն25% of recorder trees were about The minimum and maximum ®re-free intervals and the twice as long as values computed for all recorded ®res, Lower and Upper Exceedance Intervals all decreased ranging from 29% to 130% increases in the ®re-free with decreasing elevation (Table 2). The minimum ®re- intervals (Table 4). This suggests that many ®res in the free interval ranged from 1 yr at low to mid-elevation low elevation sites were small and localized. In con- sites to 5±6 yr at higher elevation sites. The LEI ranged trast, MEI values for the three high elevation sites from just 2 yr at low elevations to as high as 12 yr at showed lower differences between all ®res and wide- higher elevations. Maximum ®re-free intervals ranged spread ®res (between 22% and 36% increases), sug- from 18±19 yr in low to mid-elevation sites to 50±51 gesting that most ®res in these mixed conifer stands yr at higher elevation sites. The UEI ranged from 10± burned through most of the stand. This observation is 12 yr at lower elevations to as high as 38±50 yr at clearly seen in the composite graphs (Fig. 2). Many upper elevation sites. more, isolated, local ®res occurred in the lower ele- The variability in ®re-free intervals is itself a func- vation pure ponderosa pine stands than occurred in the tion of both the range between the minimum and max- high elevation stands. imum intervals and the regularity of years between Fire-free intervals at SMI, which has potential bar- successive ®res. No clear elevational pattern, however, riers to ®res spreading into the area, were generally was evident in the variability of ®re-free intervals. longer than at the other low and mid-elevation sites High-elevation sites had both high variability (CV ϭ that had similar vegetation but no topographic barriers. 0.73 at MNT) and low variability (CV ϭ 0.43 at TCK), Most measures of central tendency in ®re-free intervals similar to the range of values at low-elevation sites. at this site are considerably longer than the other low- Variability was highest at site SMI (CV ϭ 0.84), con- elevation ponderosa pine sites (Table 2), while the MEI June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1715

FIG. 3. Distributions of the ®re-free intervals for all nine study sites during the reference period, and for all nine sites combined (All). The probability density function, f(t), models the ®re-free interval data based on the Weibull distribution, and can be used to assess the goodness of ®t. The hazard rate is represented by h(t). Steep slopes indicate sites with shorter ®re-free intervals and greater degree of ®re hazard. 1716 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6

TABLE 2. Descriptive statistics for ®re-free interval analyses for nine study sites in the San Juan Mountains, arranged in order of decreasing elevation (from left to right).

Statistic² BCN TCK MNT SMI BCK HCK TRS PLT FPC All³ Begin year§ 1729 1748 1654 1685 1729 1679 1707 1645 1703 1650 Total no. 5 7 11 15 20 19 16 38 26 91 intervals Mean ®re 30 19 21 13 8 11 11 6 7 3 interval (yr) Median interval 2919219899652 (yr) MEI (yr) 28 18 18 10 7 9 10 6 6 2 MOI (yr) 23 18 10 3666442 Minimum 6 542112111 interval (yr) Maximum 51 28 50 32 19 27 31 18 20 9 interval (yr) LEI (yr) 12 10 63333221 UEI (yr) 50 27 38 25 13 19 19 10 12 5 CV 0.59 0.43 0.73 0.84 0.57 0.66 0.71 0.59 0.69 0.71 Skewness Ϫ0.16 Ϫ0.51 0.61 0.70 0.64 0.81 1.30 1.11 1.29 1.39 ² Abbreviations: MEI ϭ Weibull Median Interval; MOI ϭ Weibull Modal Interval; LEI ϭ Lower Exceedance Interval; UEI ϭ Upper Exceedance Interval; CV ϭ coef®cient of variation. ³ All sites combined. § The beginning year for analysis was the ®rst year with a minimum of two trees scarred. The ending year of analysis for all sites was 1880. for more widespread ®res is over twice that found at stands, (3) differences in cambial phenology among the other ponderosa pine sites (Table 4). These results trees, or (4) our within-ring estimation process of ®re- suggest that topographic barriers have a considerable scar position is too coarse for accurate determination effect on ®re frequency and likely in¯uence local veg- of past ®re seasons. etation communities.

Fire seasonality Fire±climate associations For all nine sites combined, 57% of all ®res prior to 1880 occurred during the spring dormant season (April Fires occurred more often during drought years. This to mid-June; see Table 5). Fires were less frequent be- statistically signi®cant (P Ͻ 0.001) trend was evident tween mid-June and mid-July (only 12% of all ®res), based on all ®res regardless of extent (Fig. 6A, Year the period that corresponds to the height of the current 0) as well as on widespread ®res (recorded on 25% of summer monsoon season. These ®res likely occurred all recorder trees; Fig. 6B). Above-average rainfall in in years when monsoon rainfall either was diminished, years t Ϫ 3 and t Ϫ 2 prior to the ®re year may have or delayed, or failed altogether. Another peak in ®re facilitated ®re occurrence as indicated by indices that activity occurs in the late portion in the growing season approached the ϩ0.95 con®dence level. Although most (mid-July and after, 31% of all ®res). The relative pro- of these indices were not statistically signi®cant, their portions of dormant and growing season scars ¯uctu- overall pattern and potential in¯uence on ®re occur- ated somewhat over time, but we found no apparent rence should not be overlooked. A statistically signif- temporal trends that might indicate changing season- icant (P Ͻ 0.01) relationship was observed, however, ality of ®res. In addition, no trends were apparent along between rainfall in year t Ϫ 2 and ®re occurrence based the range of elevations represented by our sites. We on ®res that scarred at least 25% of the sampled trees did uncover, however, striking differences among study sites. For example, most ®res at the PLT site occurred (Fig. 6B). This suggests that ®res that are more wide- during the dormant season (82%), whereas summer spread are additionally in¯uenced by abundant mois- burns (mid-June through August) predominated at TRS ture in previous years, thus increasing ®ne fuel loadings (64%). We also observed that ®res during certain years and contributing to more homogeneous fuel coverage occurred prior to and throughout the growing season across the landscape in years preceding ®re occurrence. (Fig. 5). Similar ®ndings were observed for other lo- Finally, we conducted separate superposed epoch anal- cations in the Southwest (Baisan and Swetnam 1990). yses (not shown) on the 1680±1752 and 1767±1880 These multiple seasonal estimates for a single year in- sub periods to determine whether the relationship be- dicate either (1) single-ignition ®res that were persis- tween ®re and climate had changed temporally between tent throughout the growing season within individual periods. The two sets of results were very similar to stands, (2) multiple-ignition ®res that occurred during the previous analysis, indicating no change in the re- different periods within and/or among the sampled sponse by ®re to climate between periods. June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1717

FIG. 4. Fire chronologies for each of nine sites representing (A) all ®res and (B) ®res that scarred Ն25% of the sampled trees (i.e., widespread ®res). A hiatus in ®re occurrence is indicated between ϳ1750 and 1770; this is especially obvious in the lower graph. The Composite axis represents information from all sites.

DISCUSSION ni®cant, direct relationship between precipitation in the How did pre-1880 ®res vary temporally Southwest and the Paci®c Decadal Oscillation (PDO; within the San Juan Mountains? see Mantua et al. 1997). An extended series of cool- phase PDO events (i.e., a southerly ¯ow of cold We found a clear hiatus in ®re occurrence between waters off the western North American coast) could ϳ1750 and 1770, a few decades earlier than reported reduce ®re occurrence due to lower overall tempera- for other sites throughout the American Southwest. tures that would facilitate high pressure dominance, Several studies have noted a reduced amplitude of El thus inhibiting convectional uplift and minimizing NinÄo-Southern Oscillation (ENSO) variation in the late lightning activity. These cooler climatic characteristics 1700s to early 1800s (Cleaveland et al. 1992, Swetnam could also lower evaporation rates, thus allowing fuels and Betancourt 1998, Veblen et al. 2000). A decrease to retain moisture longer and reducing the probability in ENSO activity would diminish winter/spring rainfall of successful ®re ignitions. Finally, summer rainfall in the Southwest, reducing ®ne fuels necessary for suc- and thunderstorm activity may have been diminished cessful ®re ignition and spread. In addition, a recent study (Grissino-Mayer et al. 2002) demonstrated a sig-

TABLE 4. Comparison of Weibull Median Interval (MEI) for all ®res and for ®res that scarred Ն25% of the recorder TABLE 3. Differences in mean ®re-free intervals between trees (a minimum of two trees scarred) at each of our nine the 1680±1752 and the 1767±1880 periods for all ®re years study sites in the San Juan Mountains, arranged from high as well as for years when Ն25% of the sampled trees were elevation to low elevation. scarred per site, with a minimum of two ®re scars per year to indicate more widespread ®res. MEI (yr) Fire type and No. Mean Study site All Ն25% Increase (%) period intervals interval² df tP BCN 28 38 36 All ®res TCK 18 22 22 1680±1752 21 3.2 72 2.03 0.04 MNT 18 22 22 1767±1880 53 2.1 SMI 10 23 130 BCK 7 9 29 Widespread ®res HCK 9 14 56 1680±1752 15 4.5 55 2.16 0.03 TRS 10 13 30 1767±1880 42 2.7 PLT 6 9 50 FPC 6 11 83 ² Data have been normal transformed. 1718 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6

TABLE 5. Fire scar formation during different portions of the growing season during the pre-1880 reference period, in nine study sites.

No. ®re scars (% of total for season) Time of year² BCN TCK MNT SMI BCK HCK TRS PLT FPC All sites D: April to mid June 7 (50) 29 (72) 16 (31) 26 (42) 61 (62) 53 (47) 31 (36) 142 (82) 31 (52) 396 (57) EE: mid June to early July 0 (0) 3 (8) 2 (4) 4 (6) 4 (4) 5 (4) 6 (7) 5 (3) 7 (12) 36 (5) ME: early July to mid-July 0 (0) 0 (0) 4 (8) 1 (2) 15 (15) 18 (16) 6 (7) 0 (0) 7 (12) 51 (7) LE: mid July to late July 6 (43) 6 (15) 16 (31) 10 (16) 15 (15) 26 (23) 35 (40) 9 (5) 7 (12) 130 (19) L: late July and August 1 (7) 2 (5) 13 (26) 21 (34) 3 (3) 10 (9) 9 (10) 18 (10) 7 (12) 84 (12) ² See Methods: Fire seasonality for detailed explanation. if the mechanism that affected ENSO/PDO-driven pre- alously strong high pressure existed over the eastern cipitation also affected summer monsoon activity. Gris- Paci®c near British Columbia during the 1600s, early sino-Mayer and Swetnam (2000) hypothesized that 1700s, and 1900s. This blocking ridge would facilitate summer monsoon rainfall increased signi®cantly after meridional circulation patterns in the interior western the hiatus as shown by the changing seasonality of ®res by de¯ecting zonal ¯ow from the Paci®c in northwestern New Mexico. Any one or a combi- to the north and/or south. This high pressure region is nation of these mechanisms could have caused a hiatus consistent with cold ocean waters located in the eastern in ®re occurrence. Paci®c that could have resulted from an anomalously We further propose a possible fourth mechanism. strong series of cool-phase PDO events. Furthermore, Changes in air mass boundaries may have affected the the long gap in ®re occurrence is more suggestive of general climate characteristics of the region and there- an in¯uence by the PDO, which operates on decadal fore affected ®re occurrence. This hypothesis is sup- time scales, compared to the frequency of 3±5 yr for ported by the temporal variation in the onset of this ENSO activity. Such changes would alter the intraan- gap in ®re occurrence at various locations throughout nual distribution of frontal and convectional precipi- the western United States. This suggests a possible tation and the probability of successful lightning ig- north-south spatial pattern. Veblen et al. (2000) re- nitions, thus altering ®re regimes. The initial in¯ux of ported a gap from 1653 to 1779 in northern Colorado; this new cooler wintertime air mass could have been Brown et al. (1999) found a gap between 1723 and a severe change to more southerly ecosystems, enough 1851 in central Colorado; and here we report a gap perhaps to cause a near complete hiatus in ®res, after between 1750 and 1770 in the southern San Juan Moun- which ®res would resume, but with different properties tains. Grissino-Mayer and Swetnam (2000) reported a due to the climatically altered ®re regime. gap from 1780 to 1800 in northwestern New Mexico; Widespread ®res in the San Juan Mountains coincide and sites in the central and southern Southwest showed with large ®res elsewhere in the Southwest, which fur- a gap from 1800 to ϳ1850 (Swetnam and Baisan 1996). ther supports the hypothesis that ®re occurrence in the There is no reason to assume that the climatic factor(s) San Juan Mountains was in¯uenced by changes in cli- responsible for creating this long gap began at exactly mate. Eight of the 11 years in which three or more of the same time and had a similar duration and effect at our nine sites recorded widespread ®res (Ն25% of re- all locations. The responsible climate factor(s) may corder trees scarred) were also among the top 20 ®re have been present in the northern portion of the South- years in New Mexico and Arizona (i.e., 1685, 1724, west (including Colorado) earlier than in locations that 1729, 1748, 1842, 1851, 1879, and 1880; Swetnam and are more southerly. This suggests an expansion of a Baisan 1996). Two of the major ®re years in the San dominant air mass from a more northerly location and/ Juan Mountains (1786 and 1851) were also major ®re or strengthened in¯uences of regional-scale climatic years in northern Colorado (Brown et al. 1999, Veblen phenomena down into the San Juan Mountains and et al. 2000). eventually farther south. The shift in ®re regimes in the San Juan Mountains Particularly relevant is the winter air mass boundary following the 1750±1770 hiatus is strongly indicated that affects the Paci®c Northwest, which has its south- by changes in ®re-free intervals. Prior to 1750, ®re ern boundary along the northern Colorado±Utah±Ne- occurrence was generally low, but was higher after vada border (Mitchell 1976). Based on tree-ring data 1770. The effect of the 1748 ®re must also be consid- from the Great Basin region, Woodhouse and Kay ered a contributing factor. Because the 1748 ®re was (1990) determined that a meridional ¯ow pattern (i.e., particularly widespread, occurring in eight of our nine north to south) likely dominated during the 1600s, early study sites, it may have created a landscape in the San 1700s, and in the 1900s, while a zonal ¯ow orientation Juan Mountains with more homogeneous fuel charac- (i.e., east to west) dominated during the late 1700s and teristics. Fires that occurred after the long hiatus would 1800s. A reconstruction of climate based on a network therefore be expected to be widespread ®res, and this of tree-ring sites (Fritts et al. 1979) revealed that anom- is indeed the case. A period after the hiatus (1767± June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1719

occurred earlier by allowing them to become more widespread. Further, our results indicate widespread ®re occurrence is likely in years when (1) spring/sum- mer rainfall amounts are well below average, (2) rain- fall during preceding years was above average, and (3) spring/summer rainfall two years prior was exception- ally heavy. This scenario is enhanced if a particularly wet El NinÄo event is succeeded by a particularly dry La NinÄa event in months leading up to the ®re event. We found few ®re events between 1890 and 1989, all of which were very limited in extent (generally a single ®re scar in a single stand), a pattern that is con- sistent with ®ndings throughout the western United States. Between ϳ1890 and 1930, only two ®re scars (1904 and 1911) were found on two of the 94 recorder trees for that period. Fires were simply unable to spread over large areas because of changes in understory fuel conditions (initially by grazing livestock) and active ®re suppression (Covington and Moore 1994). This near absence of ®res is particularly troubling because these long ®re-free intervals far exceed both the max- imum ®re-free intervals and the Upper Exceedance In- tervals based on pre-1880 ®re regimes. The situation is more problematic in the pure ponderosa pine stands at low elevations where pre-1880 maximum intervals and Upper Exceedance Interval values were much shorter. Given that over 100 years of fuels have built up in most locations within the San Juan Mountains, any future wild®re will likely have a greater intensity (i.e., heat output) than pre-1880 ®res, have greater ef- fects on ecosystems within the San Juan Mountains across all elevational and longitudinal gradients, and be more widespread due to the increased contiguity of fuels across the landscape. This suggestion was re- cently supported by the spatially extensive Missionary Ridge ®re in June and July 2002, 17 km northeast of Durango, Colorado, which burned Ͼ29 500 ha in pon- FIG. 5. Fire seasonality during four ®re years within the San Juan Mountains. While dormant-season ®res predomi- derosa pine and mixed conifer forests. nated, the range of seasons indicates that ®res during these years probably occurred throughout the growing season. See How did pre-1880 ®res vary spatially Methods: Fire seasonality for explanations of seasonal des- within the San Juan Mountains? ignations. Each of the nine sites had a more-or-less unique ®re history (i.e., different ®re years) indicating that most 1880) had a higher percentage of ®res at two or more ®res were relatively localized to one geographic area sites. Speci®cally, there were multi-site ®res in 32 of in the San Juan Mountains (although they may have 59 ®re years (54%) in this period compared to multi- burned most of an individual stand). Widespread ®res, site ®res in 14 of 33 ®re years (42%) in an earlier period however, occurred during 11 years between 1680 and from 1650 to 1750. Thus, the post-hiatus ®re regime 1880 when ®res were recorded in three or more of our was characterized by ®res that were more frequent and study sites. These must have been severe ®re years, were generally more widespread. when much of the mountain range was ablaze. For ex- Fire occurrence is also driven by climate on shorter ample, a major widespread ®re occurred in 1748 at monthly time scales, primarily by rainfall during July. eight of our nine study sites, indicating a ®re that likely Summer monsoon rainfall begins at or near the begin- occurred throughout the entire mountain range because ning of July, and onset, failure, or possible delay is BCK is our easternmost site while FPC is our west- especially pivotal for ®re occurrence. Below average ernmost site, a distance of ϳ180 km. Given the spatial rainfall during July (i.e., a failure or delay) would lower extent of these ®res, both within the San Juan Moun- fuel moisture levels and aid successful ®re ignition and tains and throughout the Southwest, it is unlikely such spread, or enhance the severity of any ®res that had widespread ®res were set by Native Americans. 1720 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6

FIG. 6. Results from the superposed epoch analysis, comparing ®re occurrence based on (A) all ®re scars and (B) only those that scarred Ն25% of the sampled recorder trees. Year 0 is the year of ®re. The tree-ring index (mean ϭ 1.0) is a surrogate for spring rainfall. The period analyzed extends from 1654 to 1880. Total ®re years ϭ 80; widespread ®res ϭ 59. Con®dence intervals are represented by 0.95, 0.99, and 0.999.

Fire-free intervals generally were longer at higher between the Lower and Upper Exceedance Intervals) elevations. With the exception of one anomalous site was 8±16 yr in pure ponderosa pine stands at lower where topography played a key role (SMI), Weibull elevations (excluding the anomalous SMI site), and 17± Median Interval values for widespread ®res (Ն25% of 38 yr in ponderosa pine±mixed conifer stands at higher recorder trees scarred) in the pure ponderosa pine elevations. Fires at the shorter intervals, especially if stands at lower elevations were 9±14 yr, compared with they occurred during the growing season, would create 22±38 yr for ponderosa pine±mixed conifer sites at patches of mineral soil and reduce the density of com- higher elevations. The longer ®re-free intervals at high- peting shrubs such as Gambel oak (Harrington 1985, er elevations likely re¯ect the generally mesic condi- 1987), thus enhancing the germination environment for tions that occur at higher elevations caused by lower pine seedlings. Fires at short intervals, however, would temperatures that inhibit evapotranspiration, enhanced also kill young pine seedlings, thus requiring occa- precipitation from orographic uplift, and greater snow sional long ®re-free intervals to allow the seedlings to pack and increased runoff from snowmelt. These ef- grow large enough to survive ®re (Pearson 1950, White fects are more pronounced on north-facing high-ele- 1985). Fire-free intervals in the lower elevation pon- vation slopes where evapotranspiration rates are lower derosa pine forests were generally longer for wide- and fuel moisture levels are generally higher. spread ®res (Ն25% trees scarred) than for ®res of any Variability in ®re-free intervals is as important as size. This indicates that many of the lower elevation central tendency for interpreting functional implica- ®res were small and localized. Therefore, patches tions of the ®re regime, such as establishment of pon- would be present that escaped ®re long enough for new derosa pine seedlings. At all nine sites in the San Juan pine saplings to become established, even though the Mountains, the range of intervals (i.e., the difference stand as a whole was burning frequently. At the higher June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1721

TABLE 6. Comparison of ®re history statistics (in years) among ponderosa pine forests in Arizona, New Mexico, northern Colorado, and the San Juan Mountains, Colorado.

Central, Jemez San Juan Northern southern Mountains, Mountains, Front Range, Statistic Arizona² New Mexico³ Colorado§ Colorado࿣ Median ®re interval, all ®res, all stands 1±8 3±16 5±29 ´´´ Median ®re interval for Ն10% scarred, all stands 3±12 6±23 7±39 ´´´ MEI for Ն10% scarred, low elevation stands 3±14¶ 6±20# 8±13²² 12 MEI for Ն10% scarred, high elevation stands 3±9¶ 7±24# 21±38 16±31 Note: MEI ϭ Weibull Median Interval. ² Swetnam and Baisan (1996:21±22), 24 sites total. ³ Touchan et al. (1996:39) and Swetnam and Baisan (1996:21±22). § This study. ࿣ Veblen et al. (2000). ¶ Low elevation ϭ 13 sites; high elevation ϭ 11 sites. # Low elevation ϭ 8 sites; high elevation ϭ 7 sites. ²² Excludes the anomalous SMI stand (see Methods: Site selection). elevations, ®re-free intervals were generally longer Mountains (Table 6). The statistics share similar char- than at lower elevations, and most ®res at higher ele- acteristics with those from both northern Colorado and vation were extensive. northern New Mexico. Median ®re-free intervals from Our results also indicate longer ®re-free intervals the Jemez Mountains of northwestern New Mexico, occur in sites having topographic barriers to ®re spread, ϳ100 km to the south of our study area, are generally although our conclusions are tentative because we shorter except for the comparable range of values for could only assess this for one site (SMI). The Weibull the low elevation stands. MEI values from the northern Median Interval value for extensive ®res (Ն25% of Front Range for both lower and higher elevation sites recorders scarred) in the lower elevation SMI site was correspond closely with our values from pure ponde- 23 yr, compared with MEI values of 9±14 yr in all rosa pine and ponderosa pine±mixed conifer stands, other lower elevation sites. Fire-scarred trees are often respectively. rare on steep slopes above cliffs and in other locations Fire-free intervals in the San Juan Mountains, there- that appear to be somewhat protected from ®re. Madany fore, are intermediate between those reported from Ar- and West (1983) also documented very long ®re-free izona and New Mexico and from northern Colorado. intervals in a ponderosa pine forest located on a small, The reason for any differences is perhaps related to the isolated mesa top in southern Utah. Additional research different fuel complexes in the different areas. Most of is required, however, on spatial patterns of individual the Arizona and New Mexico sites have understory ®re events and the effects of local topographic variation dominated by grasses, which dry quickly and carry on ®re history in mountainous areas. surface ®re readily (predominantly the Pinus ponde- We found no compelling evidence of longer ®re-free rosa±Bouteloua gracilis or P. ponderosa±Festuca ar- intervals in the generally wetter eastern portion of the izonica habitat types; Alexander 1985). In contrast, San Juan Mountains. Of the lower elevation sites, MEI shrubs are generally more important than grasses in the values for extensive ®res (Ն25% of recorder trees understory of pine forests in southwestern Colorado scarred) were 9 yr and 11 yr in the two western sites (the Pinus ponderosa±Quercus gambelii habitat type; compared with 9 yr and 13 yr in the two eastern sites. Alexander 1985, Floyd-Hanna et al. 1996). The shrubs For higher elevation sites, MEI values were 22 yr and dry more slowly than do the grasses, and may generally 38 yr in two western sites compared with 22 yr in one support a slower rate of ®re spread. eastern site. Conclusive evidence of a longitudinal gra- dient in ®re regimes would require additional ®re his- Did Native Americans in¯uence tory sites peripherally located in the San Juan Moun- pre-1880 ®re regimes? tains. The TRS site is a well-documented prehistoric Ute encampment settlement where Native Americans may With which region is the San Juan Mountains have augmented the natural ®re frequency and altered ®re regime more aligned? the seasonality of past ®res with intentionally set ®res. Direct comparisons of ®re history statistics from dif- However, measures of central tendency were all longer ferent studies are somewhat dif®cult because of dif- (6±11 yr) than at any of the other lower elevation pon- ferences in the statistics used, sampling intensity, and derosa pine sites (4±8 yr) except the anomalous SMI sizes of the study areas. Three recent compilations of site. This ®nding does not support the hypothesis that ®re history statistics from southern/central Arizona, ®re-free intervals would be shorter in an area where northern New Mexico, and northern Colorado, how- ®re ignition was likely to be augmented by Native ever, allow cautious comparisons with the San Juan Americans. The TRS site, however, has both a lower 1722 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6 proportion of dormant season ®res (36%) and a higher about the general ecology of the San Juan Mountains. We percentage of late season ®res (50%) than any of the thank our ®eld assistants, Lisa Bartlett, Dustin Hanna, Matt Romme, Dan Tinker, Sally Tinker, Joe Vanlandschoot, and other lower elevation sites. Augmentation of late sea- Owen Williams. We also thank Rex Adams, Laura Baxter, son ®res by Native American burning could result in and Jim Fairchild for providing us with the Rampart Hills fewer dormant season ®res. Nonetheless, this result North tree-ring data. Will Fontanez and Will Arbaugh, Car- alone does not provide convincing evidence that Native tographic Services Laboratory, University of Tennessee, drafted Fig. 1. Fort Lewis College, Harvard Forest, and the American people were augmenting ®re ignitions in the Laboratory of Tree-Ring Research (University of Arizona) vicinity of their encampment at Turkey Springs. Our provided facilities, equipment, and logistical support. Peter results, though limited to a single site, support the view Z. Fule and an anonymous reviewer provided useful com- that human ignitions were relatively unimportant in ments that greatly improved an earlier version of this man- controlling ®re history in the San Juan Mountains as uscript. Data sets for all nine sites have been permanently archived with the International Multiproxy Paleo®re Database a whole. (IMPD) at the World Data Center for Paleoclimatology in Boulder, Colorado, through NOAA's National Geophysical CONCLUSIONS Data Center.

The ®re regime of ponderosa pine forests in the LITERATURE CITED southern San Juan Mountains currently exists outside Agee, J. K. 1993. Fire ecology of Paci®c Northwest forests. its historical range of variability for ®re frequency and Press, Washington, D.C., USA. extent. This suggests that future ®res will be larger and Alexander, R. R. 1985. Major habitat types, community more severe than previous ®res because of the fuel that types, and plant communities in the Rocky Mountains. has accumulated during the anomalously long current USDA Forest Service General Technical Report RM-123, ®re-free interval. The pre-1880 ®re regime exhibited Rocky Mountain Forest and Range Experiment Station, Fort Collins, Colorado, USA. substantial spatial and temporal variability. For ex- Allen, C. D. 2002. Lots of lightning and plenty of people: ample, stands at higher elevations and in proximity to an ecological history of ®re in the upland southwest. Pages topographic barriers to ®re spread have had generally 143±193 in T. R. Vale, editor. Fire, native peoples, and the longer ®re intervals, and ®re intervals within individual natural landscape. Island Press, Washington, D.C., USA. Arno, S. F. 1980. Forest ®re history in the northern Rockies. stands varied from Ͻ5yrtoϾ30 yr. Very extensive Journal of Forestry 78:460±465. ®res occurred in past years and should be expected in Arno, S. F., and K. M. Sneck. 1977. A method for determining the future. Past years of extensive ®re generally were ®re history in coniferous forests of the mountain west. very dry years that followed two to three years of gen- USDA Forest Service Intermountain Forest and Range Ex- erally wetter conditions. Fires in the San Juan Moun- periment Station General Technical Report INT-GTR-42, Ogden, Utah. tains (and perhaps throughout much of the western Arnold, J. F. 1950. Changes in ponderosa pine bunchgrass United States) may have been directly impacted by ranges in northern Arizona resulting from pine regeneration signi®cant changes in past climate on interdecadal time and grazing. Journal of Forestry 48:118±126. scales. We believe these changes were related to weak- Baisan, C. H., and T. W. Swetnam. 1990. Fire history on a ened El NinÄo-Southern Oscillation activity, an extend- desert mountain range, Rincon Mountain, Wilderness, Ar- izona, USA. Canadian Journal of Forest Research 20:1559± ed series of cool-phase Paci®c Decadal Oscillation 1569. events, and weakened monsoonal moisture, all possibly Balling, R. C., Jr., G. A. Meyer, and S. G. Wells. 1992. Cli- entrained in an invasive air mass typical of locations mate change in Yellowstone National Park: is the drought- that are more northerly. Many studies have ®rmly es- related risk of wild®res increasing? Climatic Change 22: 35±45. tablished that the climate of the 20th and 21st centuries Barrett, S. W., and S. F. Arno. 1982. Indian ®res as an eco- is not comparable to climate during the 18th and 19th logical in¯uence in the northern Rockies. Journal of For- centuries due to human alterations of the atmosphere. estry 80:647±651. Wild®res may not function or behave during the present Bergeron, Y., and M. D. Flannigan. 1995. Predicting the ef- and future as they did under pre-EuroAmerican envi- fects of climate change on ®re frequency in the southeastern Canadian boreal forest. Water, Air and Soil Pollution 82: ronmental conditions due to the altered climate re- 437±444. gimes. Finally, although Native American people un- Bolton, H. E. 1950. Pageant in the wilderness: the story of doubtedly ignited ®res during the centuries before the Escalante expedition to the interior basin, 1776. Utah 1880, they apparently had a limited impact on the over- State Historical Society, Salt Lake City, Utah, USA. all ®re regime of the San Juan Mountains. Our ®ndings Brown, P. M., M. K. Hughes, C. H. Baisan, T. W. Swetnam, and A. C. Caprio. 1992. Giant sequoia ring-width chro- indicate that ®re frequency and extent prior to 1880 nologies from the central Sierra Nevada, . Tree- were controlled primarily by climatic in¯uences. Ring Bulletin 52:1±14. Brown, P. M., M. R. Kaufmann, and W. D. Shepperd. 1999. ACKNOWLEDGMENTS Long-term, landscape patterns of past ®re events in a mon- This research was funded by the USDA Forest Service, San tane ponderosa pine forest of central Colorado. Landscape Juan National Forest, Durango, Colorado. We thank Jim Pow- Ecology 14:513±532. ers and Thurman Wilson for helping secure funding, and Jim Brown, P. M., and C. H. Sieg. 1996. Fire history in interior Webb for making this research a high priority. We also thank ponderosa pine communities of the Black Hills, South Da- Jeff Redders, Phil Kemp, and Albert Spencer for assistance kota, USA. International Journal of Wildland Fire 6:97± in locating study sites and providing meaningful information 105. June 2004 FIRE REGIMES OF THE SAN JUAN MOUNTAINS 1723

Brown, P. M., and C. H. Sieg. 1999. Historical variability in Fisher, R. F., M. J. Jenkins, and W. F. Fisher. 1987. Fire and ®re at the ponderosa pine±northern Great Plains prairie the prairieÐforest mosaic of Devil's Tower National Mon- ecotone, southeastern Black Hills, South Dakota. Ecosci- ument. American Midland Naturalist 117:250±257. ence 6:539±547. Fleischner, T. L. 1994. Ecological costs of livestock grazing Brown, P. M., and T. W. Swetnam. 1994. A cross-dated ®re in western North America. Conservation Biology 8:629± history from coast redwood near Redwood National Park, 644. California. Canadian Journal of Forest Research 24:21±31. Floyd-Hanna, L., A. W. Spencer, and W. H. Romme. 1996. Callaway, D., J. Janetski, and O. C. Stewart. 1986. Ute. Pages Biotic communities of the semiarid foothills and valleys. 336±367 in W. L. D'Azevedo, editor. Handbook of North Pages 143±158 in R. Blair, editor. The western San Juan American Indians. Volume 11. Great Basin. Smithsonian Mountains: their geology, ecology, and human history. Uni- Institution, Washington, D.C., USA. versity Press of Colorado, Niwot, Colorado, USA. Cleaveland, M. K., E. R. Cook, and D. W. Stahle. 1992. Fritts, H. C., G. R. Lofgren, and G. A. Gordon. 1979. Var- Secular variability of the Southern Oscillation detected in iations in climate since 1602 as reconstructed from tree tree-ring data from Mexico and the southern United States. rings. Quaternary Research 12:18±46. Pages 271±291 in H. F. Diaz and V. Markgraf, editors. El Fritts, H. C., D. G. Smith, and M. A. Stokes. 1965. The NinÄo: historical and paleoclimatic aspects of the Southern biological model for paleoclimatic interpretation of Mesa Oscillation. Cambridge University Press, New York, New Verde tree-ring series. American Antiquity 31:101±121. York, USA. FuleÂ, P. Z., and W. W. Covington. 1997. Fire regimes and Cooper, C. F. 1960. Changes in vegetation, structure, and forest structure in the Sierra Madre Occidental, Durango, growth of southwestern pine forests since white settlement. Mexico. Acta Botanica Mexicana 41:43±79. Ecological Monographs 30:129±164. FuleÂ, P. Z., W. W. Covington, and M. M. Moore. 1997. De- Covington, W. W., and M. M. Moore. 1994. Southwestern termining reference conditions for ecosystem management ponderosa forest structure: changes since Euro-American of southwestern ponderosa pine forests. Ecological Appli- settlement. Journal of Forestry 92:39±47. cations 7:895±908. Dahms, C. W., and B. W. Geils. 1997. An assessment of forest Grissino-Mayer, H. D. 1999. Modeling ®re interval data from ecosystem health in the Southwest. USDA Forest Service, the American Southwest with the Weibull distribution. In- Rocky Mountain Forest and Range Experiment Station, ternational Journal of Wildland Fire 9:37±50. General Technical Report RM-GTR-295, Fort Collins, Grissino-Mayer, H. D. 2001. FHX2Ðsoftware for analyzing Colorado, USA. temporal and spatial patterns in ®re regimes from tree rings. Dean, J. S., and W. J. Robinson. 1978. Expanded tree-ring Tree-Ring Research 57:115±124. chronologies for the southwestern United States. Labora- Grissino-Mayer, H. D., C. H. Baisan, K. A. Morino, and T. tory of Tree-Ring Research Chronology Series 3, Univer- W. Swetnam. 2002. Multi-century trends in past climate sity of Arizona, Tucson, Arizona, USA. for the Middle Rio Grande Basin, AD 622±1992. Final DeVelice, R. L., J. A. Ludwig, W. H. Moir, and F. Ronco, Jr. Report, USDA Forest Service, Middle Rio Grande Basin 1986. A classi®cation of forest habitat types of northern Program, Albuquerque, New Mexico, USA. New Mexico and southern Colorado. USDA Forest Service Grissino-Mayer, H. D., and T. W. Swetnam. 2000. Century- General Technical Report RM-GTR-131, Rocky Mountain scale climate forcing of ®re regimes in the American South- Forest and Range Experiment Station, Fort Collins, Col- west. Holocene 10:207±214. orado, USA. Harrington, M. G. 1985. The effects of spring, summer, and Dieterich, J. H. 1980. The composite ®re intervalÐa tool for fall burning on Gambel oak in a southwestern ponderosa more accurate interpretation of ®re history. Pages 8±12 in pine stand. Forest Science 31:156±163. M. A. Stokes and J. H. Dieterich, editors. Proceedings of Harrington, M. G. 1987. Phytotoxic potential of Gambel oak the ®re history workshop. USDA Forest Service General on ponderosa pine seed germination and initial growth. Technical Report RM-GTR-81, Rocky Mountain Forest and USDA Forest Service Research Paper RP-277, Rocky Range Experiment Station, Fort Collins, Colorado, USA. Mountain Forest and Range Experiment Station, Fort Col- Dieterich, J. H., and T. W. Swetnam. 1984. Dendrochronol- lins, Colorado, USA. ogy of a ®re-scarred ponderosa pine. Forest Science 30: Heyerdahl, E. K., D. Berry, and J. K. Agee. 1995. Fire history 238±247. database of the western United States. U.S. Environmental Dishman, L. 1982. Ranching and farming in the lower Do- Protection Agency, Of®ce of Research and Development. lores River valley. Pages 23±41 in G. D. Kendrick, editor. Report EPA/600/R-96/081, Washington, D.C., USA. The river of sorrows: the history of the lower Dolores River Heyerdahl, E. K., L. B. Brubaker, and J. K. Agee. 2001. valley. U.S. Government Printing Of®ce, Denver, Colora- Spatial controls of historical ®re regimes: a multiscale ex- do, USA. ample from the interior west, USA. Ecology 82:660±678. DuBois, C. 1903. Report on the proposed San Juan Forest Kaye, M. W., and T. W. Swetnam. 1999. An assessment of Reserve, Colorado. Unpublished report, on ®le at the Su- ®re, climate, and Apache history in the Sacramento Moun- pervisor's of®ce, San Juan National Forest, Durango, Col- tains, New Mexico. 20:305±330. orado, USA. Madany, M. H., and N. E. West. 1983. Livestock grazing± Duke, P. 1995. Working through theoretical tensions in con- ®re regime interactions within montane forests of Zion Na- temporary archaeology: a practical attempt from south- tional Park, Utah. Ecology 64:661±667. western Colorado. Journal of Archaeological Method and Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. Theory 2:201±229. C. Francis. 1997. A Paci®c interdecadal climate oscillation Ellis, R. N. 1996. The Utes. Pages 225±233 in R. Blair, editor. with impacts on salmon production. Bulletin of the Amer- The western San Juan Mountains: their geology, ecology, ican Meteorological Society 78:1069±1079. and human history. University Press of Colorado, Niwot, Mitchell, V. C. 1976. The regionalization of climate in the Colorado, USA. western United States. Journal of Applied Meteorology 15: Everett, R., P.Hessburg, J. Lehmkuhl, M. Jensen, and P.Bour- 920±927. geron. 1994. Old forests in dynamic landscapes: dry-site Oliver, W. W., and R. A. Ryker. 1990. Ponderosa pine. Pages forests of eastern Oregon and Washington. Journal of For- 413±424 in R. M. Burns and B. H. Honkala, technical estry 92:22±25. coordinators. Silvics of North America. Volume 1. Coni- 1724 HENRI D. GRISSINO-MAYER ET AL. Ecology, Vol. 85, No. 6

fers. USDA Forest Service Agricultural Handbook 654, Swetnam, T. W., C. D. Allen, and J. L. Betancourt. 1999. Washington, D.C., USA. Applied historical ecology: using the past to manage for Pearson, G. A. 1950. Management of ponderosa pine in the the future. Ecological Applications 9:1189±1206. Southwest, as developed by research and experimental Swetnam, T. W., and C. H. Baisan. 1996. Historical ®re re- practice. USDA Forest Service Agriculture Monograph 6. gime patterns in the southwestern United States since AD Romme, W. H., D. W. Jamieson, J. S. Redders, G. Bigsby, J. 1700. Pages 11±32 in C. D. Allen, technical editor. Fire P. Lindsey, D. Kendall, R. Cowen, T. Kreykes, A. W. Spen- effects in Southwestern forests. USDA Forest Service Gen- cer, and J. C. Ortega. 1992. Old-growth forests of the San eral Technical Report RM-GTR-286, Los Alamos, New Juan National Forest in southwestern Colorado. Pages 154± Mexico, USA. 165 in M. R. Kaufmann, W. H. Moir, and R. L. Bassett, Swetnam, T. W., and J. L. Betancourt. 1992. Temporal pat- technical coordinators. Old-growth forests in the Southwest terns of El NinÄo/Southern Oscillation±wild®re teleconnec- tions in the southwestern United States. Pages 259±270 in and Rocky Mountain Regions. USDA Forest Service Gen- H. F. Diaz and V. Markgraf, editors. El NinÄo: historical and eral Technical Report RM-GTR-213, Fort Collins, Colo- paleoclimatic aspects of the Southern Oscillation. Cam- rado, USA. bridge University Press, New York, New York, USA. Seklecki, M. T., H. D. Grissino-Mayer, and T. W. Swetnam. Swetnam, T. W., and J. L. Betancourt. 1998. Mesoscale dis- 1996. Fire history and the possible role of Apache-set ®res turbance and ecological response to decadal climatic var- in the Chiricahua Mountains of southeastern Arizona. Pag- iability in the American Southwest. Journal of Climate 11: es 238±246 in P. F. Folliott, L. F. DeBano, M. B. Maker, 3128±3147. Jr., G. J. Gottfried, G. Solis-Garza, C. B. Edminster, D. G. Touchan, R., C. D. Allen, and T. W. Swetnam. 1996. Fire Neary, L. S. Allen, and R. H. Hamre, technical coordina- history and climatic patterns in ponderosa pine and mixed tors. Effects of ®re on Madrean Province ecosystems. conifer forests of the Jemez Mountains, northern New Mex- USDA Forest Service General Technical Report RM-GTR- ico. Pages 33±46 in C. D. Allen, technical editor. Fire ef- 289, Fort Collins, Colorado, USA. fects in Southwestern forests. USDA Forest Service Gen- Smith, D. A. 1982. Valley of the river of sorrows: a historical eral Technical Report RM-GTR-286, Fort Collins, Colo- overview of the Dolores River valley. Pages 9±21 in G. D. rado, USA. Kendrick, editor. The river of sorrows: the history of the Veblen, T. T., T. Kitzberger, and J. Donnegan. 2000. Climatic lower Dolores River valley. U.S. Government Printing Of- and human in¯uences on ®re regimes in ponderosa pine ®ce, Denver, Colorado, USA. forests in the Colorado Front Range. Ecological Applica- Smith, D. A. 1996. The miners: ``They builded better than tions 10:1178±1195. they knew.'' Pages 234±246 in R. Blair, editor. The western White, A. S. 1985. Presettlement regeneration patterns in a San Juan Mountains: their geology, ecology, and human southwestern ponderosa pine stand. Ecology 66:589±594. Wimmer, R. 2002. Wood anatomical features in tree-rings as history. University Press of Colorado, Niwot, Colorado, indicators of environmental change. Dendrochronologia USA. 20:21±36. Swetnam, T. W. 1990. Fire history and climate in the South- Woodhouse, C. A., and P. A. Kay. 1990. The use of tree-ring western United States. Pages 6±17 in J. S. Krammes, tech- chronologies to show spatial and temporal changes in an nical coordinator. Effects of ®re management of South- air mass boundary. Physical Geography 11:172±190. western natural resources. USDA Forest Service General Wotton, B. M., and M. D. Flannigan. 1993. Length of the Technical Report RM-GTR-191, Fort Collins, Colorado, ®re season in a changing climate. The Forestry Chronicle USA. 69:187±192.