Final Report: Upper Navajo River Valley Fire regime of a mixed-conifer forest in Southwestern Colorado

Transcription

1 Final Report: Upper Navajo River Valley Fire regime of a mixed-conifer forest in Southwestern Colorado Carissa F. Aoki 1, Peter M. Brown 1,2, William H. Romme 1 April Graduate Degree Program in Ecology and Department of Forest, Rangeland, and Watershed Stewardship, Colorado State University 2 Rocky Mountain Tree-Ring Research

2

3 EXECUTIVE SUMMARY Studies of historical fire regimes, i.e., the frequency and ecological effects of wildland fires at specified times in the past, have taken place all over the western states, with interesting and sometimes conflicting results. We now understand that different forest types have different fire regimes, and require different management strategies. Mixed-conifer forests, particularly of the mesic variety, are found throughout Banded Peak Ranch (BPR) and its sister ranches, yet this is one of the forest types about which we know little in terms of historical fire regime. Our study sought to identify fire s historical range of variability (HRV) on the property, with an eye toward informing management decision-making. Previous studies have established that Euro-American settlement marked a disruption of natural fire processes in many western forests. With this in mind, we sought to understand the pre-euro-american settlement HRV on BPR, using a combination of fire scar and stand age structure data. Because Euro-American settlers began producing significant ecological impacts in the San Juan Mountains beginning in the late 1800s, the historical time period covered by our analysis was from approximately 1700 through Some key findings: The majority of sampled trees were less than 150 years old, a legacy of logging and disturbance history. The majority of the mixed conifer forest on the ranch is of the mesic (cool and moist) variety. Fire scars were uncommon, and few of those we found represented more than a single historical fire, indicating that the spatial extent of a low-severity/high-frequency fire regime was small and that surface fire in the mesic mixed-conifer was rare. Fire scar dates provide the strongest evidence of ecologically significant past fires (i.e., fires that burned more than a single lightning-hit tree or tiny patch of forest) when they appear on multiple trees in one study area, or when they are correlated to dates found in other study areas nearby; years where one or the other was the case at BPR were: 1748, 1806, 1851, 1861, 1879, and Fire in 1879 affected a large portion of the ranch; it was also reflected in patches of aspen recruitment across the ranch, suggesting that at least part of the burned area in that year burned at high-severity. High-severity patches vary widely in both size and frequency. There is evidence of a mixed severity fire regime in a number of plots throughout the ranch. High- and mixed-severity fire occurred naturally on the landscape in the centuries prior to Euro-American settlement, so it would not be surprising or historically unprecedented for fires of this type to occur today. Fuel reduction treatments now underway in the vicinity of buildings and other ranch infrastructure will help to reduce damage from 1

4 future large, severe fires. Prescribed surface fires, such as those used for restoration in lowseverity ponderosa pine systems, probably would not be effective at improving ecological conditions in the mesic mixed-conifer forests that cover most of BPR; these latter forests are ecologically very different from lower-elevation ponderosa pine forest ecosystems. However, wildland fire use and prescribed crown burning (pending current experiments in the use of the latter) could be considered for the mesic mixed-conifer forests of BPR and adjacent areas. Forests adapted to mixed- and high-severity fire generally regenerate easily following fire, with aspen re-sprouting from surviving root systems and trees from adjacent lightly burned and unburned patches providing a seed source for conifer species. Future aspen regeneration, however, may require special consideration due to the large herd of elk now on the ranch. A drier and warmer future climate may increase fire frequency and the patch size of severely burned areas, in which case seed sources would be of greater concern. 2

5 TABLE OF CONTENTS I. INTRODUCTION...5 II. BACKGROUND...6 III. STUDY AREA...7 IV. FIELD METHODS...9 V. RESULTS General tree information...11 a. So how old are all those trees?...11 b. Are there any old-growth trees left?...12 c. What about age-size relationships?...13 d. What tree species make up the mixed conifer forests? Fire history reconstruction: Solving the mystery of historical fire regime...15 a. Clue No. 1: Fire scar data...16 b. Clue No. 2: Aspen establishment dates across the ranch...18 c. Clue No. 3: Species composition and cohort age data within plots...19 VI. DISCUSSION Management Recommendations Some additional thoughts...28 VII. REFERENCES...31 VIII. APPENDICES

6 4

7 I. INTRODUCTION Following unusually severe fire seasons throughout the western United States in 2000 and 2002, land managers and policy makers began implementing a series of fire policies aimed primarily at fuels reduction (Franklin and Agee 2003, Veblen 2003, Stephens and Ruth 2005) 1. These policies were predicated on the assumption that 20 th century fire suppression caused an unnatural build-up of fuels in many fire-dependant ecosystems, leading to the extremely large and severe fires of recent decades. Current forests, in other words, are too dense to maintain their historical low- or mixed-severity burning, and are now outside their naturally occurring patterns of both fire behavior and fuel and forest structures. (These patterns are referred to as the historical range of variability, or HRV.) This idea, however, resulted from the coincidence of the need for science on which to base policy decisions, with the fact that many of the initial fire regime studies in the west were carried out in ponderosa pine ecosystems. More recent studies in other forest types, such as higher elevation subalpine forests, have shown that their HRV may feature naturally dense stands, high-severity (stand-replacing) fires, and much longer time intervals between fires. Thus, these forests may not be outside of their HRV (e.g., Schoennagel and others 2004, Sibold and others 2006), and management actions designed for ponderosa pine forest fire regimes do not apply. In recent years, studies in these high-elevation subalpine systems have diversified our knowledge of fire regimes, and increased our understanding of how management might respond differently in these forest types. A network of fire history studies now exists throughout the western United States, extending from southern British Columbia to northern Mexico. This network encompasses a wide latitudinal range, elevational gradients, and the shift in weather pattern that occurs over the Continental Divide. The studies include a strong body of work in both the ponderosa and subalpine systems, but studies in the mixed-conifer are rare. Indeed, even a simple definition of what constitutes the mixed-conifer forest has been elusive. Dieterich (1983) pointed out that mixed-conifer may indicate a wide variety of forest types, ranging from mixed-conifer-pinegrass ecosystems in Oregon, to sequoiamixed-conifer forests in the Sierra Nevada, to ponderosa pine-white fir forests in Crater Lake National Park. Within Colorado s San Juan Mountains, Romme and others (2009) distinguished between warm-dry mixed-conifer forests, and cool-moist or mesic mixed-conifer forests, which differ from one another in elevation, aspect, major species, disturbance regime, and stand structure (see Appendix A). Cool-moist and mesic are often used interchangeably. We will hereafter use the term mesic, as this term is commonly used in most vegetation classifications. For fire management purposes, each type of mixed-conifer forest must be characterized within its own climatic, vegetative and geographic context. Banded Peak 1 References are listed at the end of this report. 5

8 Ranch and its sister ranches 2 provide a unique setting in which to study the historical fire regime of mesic mixed-conifer forests in southwestern Colorado. Both warm-dry and mesic mixed-conifer forests are found on the ranch, but mesic mixed-conifer forest is by far the more extensive. Research has been conducted on the fire ecology of warm-dry mixedconifer forests in the San Juan Mountains (see Background section below), but the mesic mixed-conifer forests have received very little study. A better understanding of how fire functions in this system will help us understand how fire might have operated historically throughout the range of forest types on the property. In higher elevations of the San Juans, mesic mixed-conifer forests occur between elevations of approximately 2200 and 3100 m (7200 to 10,100 feet). Tree species may include Douglas-fir (Pseudotsuga menziesii), Engelmann spruce (Picea engelmannii), subalpine fir (Abies lasiocarpa), white fir (Abies concolor), and quaking aspen (Populus tremuloides). Blue spruce (Picea pungens) and southwestern white pine (Pinus strobiformis) also occur in some stands. Ponderosa pine (Pinus ponderosa) may be present in small numbers, but this species is more typically associated with warm-dry mixed-conifer forests (Romme and others 2009). The objectives of our study are 1) to determine the fire regime of the Banded Peak Ranch mesic mixed-conifer forests prior to human influence caused by Euro-American settlement; and 2) to ascertain in what ways this forest type is within or outside of its historical range of variability and to formulate management recommendations based on this information. To accomplish these goals, we collected field data on the ranch in the summer of 2007 with an eye toward answering the following questions: How frequent were past fires, what was the spatial extent and distribution of highfrequency/low-severity fire relative to that of low-frequency/stand-replacing fire, and how, if at all, has this changed through time? How does the mesic mixed-conifer fire regime present at the Ranch fit with what is known of fire regimes in lower elevation ponderosa pine forests and higher-elevation spruce-fir forests? Is there a basis for defining a mixed-severity fire regime in mesic mixed-conifer forests based on this information? II. BACKGROUND A fire regime is comprised of a number of different measures of fire. The ones we will refer to most frequently here are fire severity and fire frequency. Fire severity refers to the amount of vegetation killed by the fire a low-severity fire damages trees, but most often does not kill them, whereas a high-severity fire generally kills all or most of the trees in its path. For this reason a high-severity fire is often referred to as a stand-replacing 2 Unless otherwise specified, we will hereafter refer to the three ranches together as Banded Peak Ranch. 6

9 fire. Fire frequency is most often referred to in terms of fire interval, or the average length of time between fires at any given location. A high frequency fire regime is characterized by a small number of years between fire (fire happens more often), whereas a low frequency fire regime is characterized by a greater number of years between fire (fire happens less often). In the broadest terms, two main fire regimes have been identified in fire regime studies throughout the western states: a) low-severity/high-frequency (usually in lower elevation forests such as ponderosa pine); and b) high-severity/low-frequency (usually in higher elevation forests such as subalpine spruce-fir). Management techniques typically range from relatively aggressive restoration techniques such as thinning and prescribed burning in the former, to leave-it-alone policies in the latter. Extensive research on historical fire regime has occurred in both the Front Range of Colorado and northern Arizona and New Mexico (hereafter referred to as the Southwest ). The region of southern Colorado between the Front Range and the Southwest has been less well-studied. Henri Grissino-Mayer and others (2004) have pointed out that the San Juan region bears further study because it may bridge the gap between fire regimes in the Southwest and those found in the Colorado Front Range. The only other existing studies of fire regime in the immediate area are Wu (1999) and Brown and Wu (2005). While two of these studies included a mixed-conifer component (all were focused on forests where ponderosa pine was either a sole dominant or a principal overstory species), less attention was given to the mesic end of the mixed-conifer spectrum. Studies of mixed conifer forests nearby, but not directly in the San Juans, include Margolis and others (2007) and Touchan and others (1996). At the crux of all of these studies is the influence of Euro-American settlement, which disrupted historical fire regimes toward the end of the 19 th century. The introduction of cattle and sheep grazing removed fine fuels that had previously contributed to low-severity, highfrequency fires, and a few decades later, federal fire exclusion policies extinguished many fires before they could perform their natural function in the ecosystem. Managementoriented fire research seeks to understand the pre-euro-american settlement fire regime, so that we can understand how the regime may have changed due to anthropogenic influences, and, if necessary, try to restore the conditions to which the ecosystem is naturally adapted. III. STUDY AREA Our study area included the entire property, i.e., the total of the three individual sections: Banded Peak Ranch, Catspaw, and Navajo Headwaters. Located southeast of Pagosa Springs, Colorado, with its eastern boundary on the Continental Divide, the ranch area comprises approximately 22,250 hectares (55,000 acres). The Navajo River runs down the center of the property, with its headwaters found in the upper third of the ranch. The eastern property line follows the Continental Divide, and federal land borders the property on 7

10 Temperature (Celsius) Precipitation (mm) three sides: San Juan National Forest to the west, South San Juan Wilderness to the north, and Rio Grande National Forest to the east. The nearest available climate instrumental data to compare to the mixed-conifer elevations on the ranch are from Wolf Creek Pass (3243 m, ) and Pagosa Springs (2209 m, ), each located approximately 35 km from the study area. Average annual precipitation was 1152 mm and 513 mm, respectively; average maximum January temperatures were -1.0 C and 3.3 C, and average maximum July temperatures were 18.8 C and 28.4 C. Relative to other locations in the San Juan Mountains, Banded Peak Ranch is on average cooler and wetter. Figure 1 uses PRISM (2004) modeled climatic data to compare recent temperature and precipitation data among three locations in the San Juan Mountains Annual Precipitation TCK BCN BPR Annual Maximum Temperature TCK BCN BPR Figure 1. Annual precipitation and maximum temperature at three sites in the San Juan Mountains. All are at comparable elevation and all support mixed-conifer forests. Taylor Creek (TCK) and Burnette Canyon (BCN) are located at the west end of the range near Dolores, Colorado. Banded Peak Ranch (BPR) has been noticeably wetter and cooler than the sites to the west. 8

11 IV. FIELD METHODS The main objectives of our field sampling were to collect both fire-scar and age structure data. Low-severity fires often leave injuries (fire scars) on trees without killing them, thus providing a very good record of past low-severity fire history. In other forests where fires are less frequent and of higher severity, trees are often killed outright, and little or no scar evidence remains. In these forests, age structure analyses are often the only way to obtain fire history information. If the oldest trees in a stand are younger than the known life expectancy of the dominant tree species, or if most of the trees in a stand appear to be about the same age, this suggests a stand-replacing fire at some time in the past. The use of fire scar data in conjunction with stand age structures provides a more complete picture of the fire regimes that occurred in mixed-conifer forests on Banded Peak Ranch. Fieldwork took place from June through July, Using the USGS digital elevation map for the ranch, we selected areas that were between 2500 and 3100 meters (8200 and 10,200 feet), to focus on the elevations where the mesic mixed-conifer type is known to occur. Hawth s Tools software (Beyer 2004) was used to randomly select eight site locations within this elevation range. These were stratified within the administrative boundaries of the three ranch areas, so that two sites were located on Navajo Headwaters, two on Catspaw, and four on Banded Peak Ranch. Toward the end of the field season, the opportunity arose to sample an area on Catspaw that had been recently logged for standing dead Douglas-fir. Because freshly logged stumps might reveal fire scars that otherwise would not have been visible, we added an additional ninth site within the boundaries of the recent logging. In keeping with common scientific practice, the randomized locations of the eight sites were necessary so that findings could be extrapolated to the ranch as a whole. Yet, we also needed to maintain some degree of systematic sampling so that the small-scale spatial dynamics of fire would not be missed. We therefore placed a grid of five plots at each site, using the site location as the center plot and placing the other four plots in a 500 m square box around the center plot (Figure 2). This resulted in 40 plots total across the ranch, plus the five additional Catspaw plots (UTM coordinates, elevation, slope and aspect for each plot can be found in Appendix B). At each plot, the 30 trees (living or dead) nearest to plot center and 20 cm in diameter at breast height (DBH) were sampled using a power increment borer to collect core samples, and a chainsaw was used to collect Figure 2: At each random sample point (in this case, 2P3), four additional points were placed in a grid square around the center point. wedge samples from snags, downed logs, and stumps. We sampled the larger trees in the stands 9

12 ( 20 cm DBH) because we wanted to ensure that the majority of our samples dated back prior to Euro-American settlement. For each sample, collected information included: species; diameter at breast height (DBH); diameter at sampling height (DSH; 10 cm or ~6 in above ground level); whether the sample was living, stump, or log; location relative to plot center (distance and azimuth); and remnant condition (bark still present, sapwood present, or eroded). In cases where a stump was too deteriorated for collection, its presence and location were noted, along with DSH and, if possible, species. Core samples were collected from living trees at 10 cm above ground level to minimize missed rings between ground (germination) and coring heights. Fire-scar data were obtained by taking partial cross sections from living trees (a common dendrochronological method which allows the tree to live following sampling) and full cross sections from remnants, i.e., snags, downed logs, and stumps. This dead material is a bit more difficult to date, since we cannot refer to the current year for the outside ring date, as we can in the living samples. Although they are no longer living, they can give us important information about the recruitment dates of the oldest trees in that stand. If the stand is one that likely experiences a high-severity fire regime (i.e., fire that would destroy all the logs and snags in the stand), we can safely assume that the stand has not burned since those remnant trees recruited. Although logs and snags can certainly remain following standreplacing fire, such remnants would likely show evidence of fire such as charring the majority of remnants we found did not show any such evidence. Fire scars were sampled thoroughly within each plot, and additional samples were collected from a wider radius as we moved from plot to plot within a study site. Previous high-grade logging (i.e., selective removal of the largest merchantable trees) at a number of our plot locations made some of the sample collection difficult, since stumps were often too eroded for sampling. The methods described above closely follow those used by Brown and Wu in their 2005 study on an adjacent study area, both to maintain consistency among studies across different ecosystems, and to facilitate future attempts at scaling up across the San Juans. In the lab, all samples were surfaced with successively finer grits of sandpaper to enable viewing under a microscope, then visually crossdated using standard dendrochronological methods (Stokes and Smiley 1968). When a sample did not reach pith, a geometric pith estimator was used to estimate the number of years between the innermost dateable ring and the pith date. Samples that did not easily crossdate visually were further analyzed by measuring the tree ring widths using a sliding stage measurement system, then applying program COFECHA (Holmes 1983) to the ring widths to facilitate crossdating. Due to Banded Peak Ranch s cool, wet location (see Figure 1, above), many samples proved unusable due to advanced deterioration of the wood. For example, out of 834 conifer age structure samples collected in the original eight sites, 555 were dateable to a specific age, and 578 were at least partially dateable (i.e., the center of the sample was rotten or otherwise 10

13 Number of Trees undateable, but the remainder was still useful). Including aspen, 1208 samples were collected in the eight sites, and 1359 when including the ninth, non-random site. V. RESULTS 1. General tree information a. So how old are all those trees? Over the ranch as a whole, as in many other western forests, extensive tree recruitment began in the mid- to late-nineteenth century, and continued into the early 20 th century (Figure 3). In other words, the majority of sampled trees were less than 150 years old. The increase in recruitment during the late 19 th and Douglas-fir Aspen SW white pine Blue spruce Ponderosa pine Spruce Subalpine fir White fir Recruitment Year in 10-yr Bins Figure 3: Tree recruitment years for all sampled trees. early 20 th centuries has been documented throughout the Southwest, and has been at least in lower-elevation ponderosa pine forests attributed in part to an increase in moisture during this period (e.g., Savage and others 1996). The highlight of this period was a pulse of ponderosa pine regeneration apparent in age data from throughout the Southwest centered on the year The region around Banded Peak Ranch experienced a similar wet period as documented by both Palmer Drought Severity Indices (a measure of how dry the region was in any given year: Cook and others, 2004) and in modeled reconstructions of 11

14 precipitation (PRISM Group, 2004). From these two datasets, points located near Banded Peak indicate above average moisture during the first three to four decades of the 20 th century (Figure 4), which corresponds quite strongly to the peak in tree recruitment across all species. The late 20 th century was another wet period in the Southwest (Figure 4). A similar pulse of tree recruitment may have occurred during this recent wet period, but this recent pulse would not be detected by our sampling method which intentionally emphasized sampling older trees. The relative lack of trees dating to earlier periods can be attributed in part to mortality from extensive late 19 th century fires (see below), but 20 th century logging undoubtedly also played a role (as mentioned above, many of the stumps remaining in our plots could not be sampled due to decay). Douglas-fir, white fir and ponderosa pine can all live to ages upward of 300 years, so the lack of trees dating back before 1850 indicates a disturbance such as logging or fire that took out many or most of the older trees Palmer Drought Severity Index PRISM Precipitation Data 0 Figure 4. Climate history on Banded Peak Ranch. Left Y axis represents departure from the mean for the Palmer Drought Severity Index (i.e., positive numbers indicate relatively wet conditions; negative numbers indicate relatively dry); right Y axis represents annual precipitation in mm, based on PRISM data (a model derived from instrumental records collected at weather stations in the region). Note the wet period in the early 20 th century when many of the trees now present on the ranch became established. Data sources: Cook and others 2004, PRISM b. Are there any old-growth trees left Overall, as we mentioned before, the majority of individual trees are under 150 years old. Exceptions, however, can be found across the ranch. Table 1 lists the oldest living and remnant samples of each species, along with their locations. The oldest living trees that we 3 Since our sampling was not specifically targeted toward logged vs. unlogged areas, a different sampling design would be required to understand the change in overstory caused specifically by logging. 12

15 detected on BPR generally are a little younger than the oldest trees of these species found elsewhere in the San Juan Mountains, and quite a lot younger than individuals of these species elsewhere in the West (Swetnam and Brown 1992). For example, 300+ year old Engelmann spruce trees have been found in many locations in the San Juan Mountains, and 200+ year old aspen trees were found near Dolores, Colorado. White fir is an exception to this statement, in that 230 years is about the maximum life span for this species anywhere in the San Juans (Romme and others 2009). This finding, that the oldest trees on BPR are generally not as old as the oldest trees of the same species elsewhere in the region, may be a consequence of the wetter conditions of BPR (which often are associated with shorter tree life spans) or may reflect simply that our random sampling design did not take us to locations where the truly oldest trees may be growing on the ranch. In general, the older trees (i.e., those that pre-date 1879), which do not appear in clear even-aged cohorts, provide a number of different kinds of clues as to what may have occurred in terms of fire within any given stand. We discuss this further in the section below on species composition and age structure. Table 1. Table of oldest samples for each major tree species found in the mixed-conifer forest Species Oldest Remnant Pith Date* Location Douglas-fir 1634 Dolomite Lake, Plot 3 White fir 1685 Big Muddy Creek, Plot 5 Ponderosa pine Younger than oldest living Little Muddy Creek, Plot 2 Engelmann spruce Younger than oldest living Dolomite Lake, Plot 2 Aspen 1808 Bear Creek, Plot 2 Oldest Living Location Pith Date 1650 Dolomite Lake, Plot Little Muddy Creek, Plot Little Muddy Creek, Plot ** Dolomite Lake, Plot Bear Creek, Plot 5 * The pith is the center-most ring in the tree. Remnant refers to dead trees or stumps, as opposed to trees still living. ** This sample is older; the borer was not long enough to reach pith. c. What about age-size relationships? It would seem logical that the largest trees in a forest are also the oldest, but in fact this often is not the case because tree growth is so strongly influenced by local conditions, e.g., trees surrounded by many neighbors typically are substantially smaller than trees of the same age that grow in more open settings. We should also note that age and size correlations (or lack thereof) vary widely across the landscape. For example, Table 2 shows some recruitment date ranges for white fir and aspen. It is apparent from the table that the ages of the largest and smallest trees overlap considerably for both white fir and aspen. 13

16 Table 2. Comparison of oldest and youngest trees of certain size classes. Species DBH (cm) Earliest pith date (i.e., oldest tree of this size) White fir > White fir < Aspen > Aspen < Latest pith date (i.e., youngest tree of this size) Even in a single plot with relatively even-aged trees, correlations between size and age were generally low. Fortunately, tree characteristics other than size can be used to estimate relative tree age, e.g., very old ponderosa pine trees usually have very thick bark, thick branches, and a flattened crown (Huckaby et al. 2003). Unfortunately, similar characterizations have not been published for the mesic mixed-conifer species. d. What tree species make up the mixed-conifer forests on Banded Peak Ranch? Overall, the paucity of ponderosa pine across all of our plots indicates that the majority of the mixed-conifer forest on the ranch is of the mesic variety (Figure 5). Because our sampling design was random, we can extrapolate this finding to the ranch as a whole. Blue spruce 1% SW white pine 1% Overall Species Composition Ponderosa 3% Cottonwood 0% Subalpine fir 4% White fir 34% Aspen 31% Engelmann spruce 9% Douglas-fir 17% Figure 5: Percentages of tree species across sampling sites 14

17 We did find a greater amount of ponderosa pine at the non-randomly selected Catspaw site that had been recently logged, and ponderosa pine also is common on slopes near the entrance to the ranch, so we know that warm-dry mixed conifer does exist on the property. However, the fact that we needed to target an area in order to find this forest type suggests that it is not present on a widespread basis. Is it possible that Banded Peak Ranch was formerly comprised of a greater percentage of warm-dry mixed conifer, now converted to mesic due to 20 th century fire suppression? We think this is unlikely. Warm-dry vs. mesic mixed conifer is a distinction between bioclimatic types i.e., these are forest types influenced primarily by topography and climate, which in turn lead to species composition differences. A warm-dry mixed conifer forest might increase in total tree density and in the relative proportion of white fir vs. pine during a long period without fire, but it would still be categorized as warm-dry because the fundamental site conditions (climate and soils) did not change. In addition, one of the key differences between warm-dry and mesic mixed-conifer is the relative lack of ponderosa pine in the latter, and we would not expect that fire suppression would cause pine species to decline (i.e., a conversion within the last century would have still left some standing pine trees and definitely pine logs and snags, whereas there was no evidence of such pines in our plots). We note that regional climate changes may cause some stands to change over longer periods of time. For example, new data from the Front Range indicates that some ponderosa pine-dominated stands may have converted to a mixed-conifer species composition, due to high-severity late-19 th century fires, after which moist conditions in the early-20 th century favored regeneration of the mixed-conifer species over ponderosa pine (Laurie Huckaby, unpublished data). Note, however, that this represents a climatic shift, rather than human influence. See Appendix C for plot-by-plot information on current overstory species composition. 2. Fire history reconstruction: Solving the mystery of historical fire regime Because we have no way to establish fire history with absolute certainty, we use a number of proxies information that stands in for the fact that we were not there to document what happened 50, 100, or 300 years ago to piece together clues as to what the fire regime may have been on a given landscape. Fire scar data are the most obvious proxies, since they can document the exact year and location of past fires. However, these only paint a partial picture of the overall fire regime, especially past fire behavior. For example, a lowseverity fire may leave no scars at all, a scar may be completely grown over by subsequent years tree growth, or a fire that scarred one tree may not have scarred any of its neighbors. For these reasons, we hope to find similar fire years on more than one tree in a stand, or for years of widespread fire in more than one stand or even across the entire watershed. In 15

18 other words, multiple samples showing the same fire year strengthen the evidence of fire in that year. Overall, however, fire scars provide the best documentation of past fire history. The problem is that fire scars tend to occur primarily in low severity fire regimes, in which fire injures but does not kill the trees in the stand. For fire regimes that tend to kill entire stands of trees, often few or no fire scars are left behind. In these cases, stand age structure provides the next most informative proxy in helping us understand historical fire regimes. If the entire stand was killed in a past fire, evidence of that event will present itself in the form of a stand age structure in which all of the living trees established more recently than the fire that killed the trees in the previous stand. Depending on the life history of the species, these post-fire individuals may have established promptly after the fire and thus all may be of about the same age today (typical of aspen), or they may have established moreor-less continuously since the fire (typical of white fir). Another part of the difficulty with trying to glean historical information about standreplacing fires is that the fire will generally have removed any previous record of fire in that stand standing dead trees and downed logs may have been left by the last fire, but once a few decades or more have passed, these remnants will be well on their way to complete decomposition. With fire scars, a single stand may contain many scars from many different fires, from which one may calculate an average interval between fires. Using stand age structure, however, can only tell us about the most recent stand-replacing fire we must assume that the time-since-last-fire indicates something about the fire interval, but we do not have multiple examples to strengthen this assumption. Age structure data, then, require more deduction than fire scar information. In this study, we used age structure and fire scar data together, with all the different inferences contained therein, to attempt to piece together an understanding of the historical fire regime in mixed conifer over the ranch as a whole. Following are our results based on our analysis of the data, divided into three parts: fire scar data across the ranch; aspen recruitment establishment dates across the ranch; and species composition and cohort age within plots. a. Fire scar data We did not find many fire scars, suggesting that the low-severity/high-frequency fire regime type is not widespread in mixed conifer forest on the ranch. Low-severity fire regimes generally leave behind numerous fire scars, often multiple scars on a single tree. Such trees may date back several hundred years. In our sampling, however, we found few samples of this type. In addition, half the dateable fire scar samples contained only one fire scar (as opposed to multiple scars), suggesting that fires were either relatively infrequent and generally of a high enough severity to kill, rather than scar, most of the trees in each stand (the scarred trees were the rare exception). Notably, the year 1879 appeared as a fire date on nearly half the dateable fire scar samples (10 of 21 in the eight original sites, 13 of 29 including the ninth, non-random site). 16

19 1879 has been documented as a major fire year in many other studies in the region near Banded Peak Ranch, as well as across the west. Some of the regional studies that have found widespread fire in 1879 include: Grissino-Mayer and others (2004) and Wu (1999) in the San Juan Mountains, Margolis and others in the southern Sangre de Cristo Mountains (2007), and Touchan and others in the Jemez Mountains (1996). Examples from other locations in the West include: Grand Canyon National Park (Fulé et al. 2003), Rincon Mountains, Arizona (Baisan and Swetnam 1990), Rocky Mountain National Park (Sibold et al. 2006), White River National Forest (Sudworth 1899), and Animas Mountains, New Mexico (Baisan and Swetnam 1995). We found an 1879 date on trees in five of the nine sites (sites 1, 2, 3, 6 and 9; see table, Appendix D), though not at all plots in each of those sites (remembering that there are five plots per site). It is likely that these fire-scarred trees were the result of surface burning that occurred adjacent to patches of stand-replacing fire. Seven of the scarred trees had been scarred more than once during their lifetime, suggesting that a low-severity/highfrequency regime was operating in that stand, or that a mix of low-severity and high-severity fires characterized the area (note that two of these seven were located in the warm-dry site 9). But the fire scars alone cannot answer these questions about fire severity in For more information on the relative severity of the burned patches in the 1879 fire, we turned to the aspen age structure data (see next section). Also of note in the fire-scar data was a surprisingly large number of single fire scars throughout the 20 th century that dated to a variety of different years, with no one year being duplicated in any other sample (see Appendix D). In all likelihood, these represent scars that a) were actually caused by something other than fire, e.g., by lightning or by another tree falling along the trunk and scraping off the bark; or b) were caused by small fires that went out before spreading and before causing any significant change in stand structure. We made every effort not to collect samples that were obviously scarred from something other than fire (such as a skid or fell scar) or a single lightning strike that did not affect adjoining trees, but again, since many of the trees had been cut during high grading, it was often difficult to discern these details. Assuming that at least some percentage of these scars was caused by fire, it seems clear that small, non-spreading fires may have occasionally occurred in some areas of the ranch. While sampling, we observed the frequent occurrence of lightning-scarred trees that did not appear to be related to nearby patches of fire, further supporting this conclusion. Moreover, in several wilderness areas across the West, where fires have been allowed to burn without interference, it has been observed that most of the cumulative area burned in a decade is accounted for by only a few large fires, and that the majority of fires extinguish naturally before burning any substantial amount of area. This pattern results from the fact that ignition often occurs at times when fuels are too wet to carry fire. Small patches may burn but these fires go out before becoming extensive. Extensive fires generally occur in mid- to higher-elevation forests only in those rare years when ignition 17

20 combines with extreme weather conditions (wind, high temperatures, low relative humidities) to result in very continuous, extremely dry fuels. A standard suite of fire history statistics (e.g., mean fire interval) is usually computed for ponderosa pine forests where fire scars are abundant. However, because we found so little evidence of spreading, low-severity fires, it was not possible to compute these kinds of statistics for the mesic mixed-conifer forests of Banded Peak Ranch. Instead, our data indicate that extensive, ecologically significant fires have occurred infrequently throughout the past 300 years. Tiny, non-spreading fires may be ignited every year or every few years at different locations around the Ranch, but these little isolated fires extinguish naturally (or some may have been actively suppressed in recent years) before causing any detectable change in the structure or composition of the vegetation. The fires that do have an important influence on the vegetation are those that spread over a large area and burn at variable severity (including high severity see below). The intervals between these ecologically significant fires have varied through time, from a few decades to half a century, both before and after As shown in Appendix D, we documented only five different years during the past two centuries for which there was fire-scar evidence that fire burned in more than just a single isolated plot: 1956 (fire recorded in two different plots), 1902 (3 plots), 1879 (10 plots), 1851 (3 plots), and 1806 (3 plots). Three of these years (1806, 1851, and 1879) were major fire years throughout the Southwest. These also were extremely dry years throughout the region. Of all these extensive fire years in the past, the 1879 fire appears to have been the most extensive; in fact, we find legacies of 1879 in many places throughout the Ranch, e.g., in the form of aspen forests that originated just after that large fire (see below). We note that Jim Webb located numerous fire-scarred trees and stumps on the ranch in an earlier study. However, because his sampling design involved specifically searching for fire scars, rather than random sampling, his findings cannot be used to generalize over the entire ranch. Those fires likely burned only very small areas before extinguishing naturally. Webb s data-set augments ours, and is especially valuable in that it supports our observation (explained below) that very localized fires have continued to occur on BPR up until the present time. However, it is important to emphasize that the fire frequency derived from a selective inventory of this kind cannot be extrapolated to the ranch as a whole, i.e., we cannot conclude that all or even many of the fires so documented had any significant impact on the landscape. b. Aspen establishment dates across the ranch Our initial evaluation of aspen across the ranch found a notable spike in recruitment in the decade of the 1880s (Figure 6). These data support our findings in the fire-scar sampling that 18

21 Number of Trees Number of Trees Recruitment Years in 10-yr Bins Figure 6. Aspen recruitment across all study sites. 0 Site Name 19 Figure 7: Aspen recruitment during the 1880s decade, by site name. the 1879 fire burned in many locations across the ranch. The presence of even-aged aspen also indicates that many, if not most, of these patches burned at a high-severity, i.e., they were stand-replacing fires. Another interesting point about the 1880s aspen recruitment is that it was not equally distributed across the ranch. It appears that while the 1879 fire spread widely on the eastern half of the ranch and burned severely in many places, the western side was relatively unaffected. In Figure 7, we can see that sites at Bull Elk Pond, Little Muddy, Big Muddy, and Catspaw Logging (all located on the eastern side of the valley ) all included significant aspen recruitment during the 1880s, as opposed to sites at Dolomite Lake, Bear Creek, Elephant Head, and Beaver Creek (mostly on the west side), which did not. Since fire was widespread throughout the Southwest in 1879, we can conclude that climate was the primary driver of that large fire season. Species composition and cohort age data As we mentioned in the introduction to this section, age structure data provide more indirect evidence than do fire scar data. Our primary goal in using the age structure data was to draw some possible conclusions about the historical fire regime that focused on fire severity, as opposed to fire return interval. A quick overview of aspen recruitment on the ranch, described above, told us that the 1879 fire was a significant one on the ranch, and that the fires of that year burned in a number of different site locations. We then wanted to know more about the distribution of that fire year across the watershed as a whole. Did it burn in large patches, covering entire ridgetops or drainages? On a per plot basis (our smallest scale), was the fire more widely distributed on the eastern side of the ranch than the western side, as the overall aspen recruitment data seemed to show? Is there clear evidence of a mixed-severity fire regime, in which low and high-severity patches may occur in close

22 proximity to one another, not just within the same site, but perhaps within a single plot? We analyzed the age structure data of both conifer and aspen species, at the plot level, and put this information together with the fire scar data to help us determine the historical fire severity of each of our 40 plots, and then to summarize this information across the landscape. We note that there is no standard methodology for all of this, because so few previous fire history studies have been conducted in the kinds of forests that dominate Banded Peak Ranch. Therefore, we had to develop a suitable method, drawing upon and modifying methods that have been used in other study areas. See Appendix E for the details of this method. Following are our results. Table 3 below summarizes our conclusions about each plot, as well as the overall percentage of plots that contained mixed vs. high-severity regimes. We should re-state here that our definition of severity relates to the effect of fire on the stand, i.e., high-severity kills all the trees in a stand, while mixed-severity kills some, but not all trees. We then Site Plot Severity/Plot Scale Site Plot Severity/Plot Scale Dolomite Lake 1 NFI Elephant Head 1 ND Dolomite Lake 2 NFI Elephant Head 2 High Dolomite Lake 3 NFI Elephant Head 3 Mixed Dolomite Lake 4 ND Elephant Head 4 Mixed Dolomite Lake 5 Mixed Elephant Head 5 NFI Bull Elk Pond 1 High Big Muddy 1 High Bull Elk Pond 2 High Big Muddy 2 High Bull Elk Pond 3 High Big Muddy 3 High Bull Elk Pond 4 High Big Muddy 4 High Bull Elk Pond 5 High Big Muddy 5 Mixed Little Muddy 1 Mixed Johnson Mtn 1 High Little Muddy 2 Mixed Johnson Mtn 2 High Little Muddy 3 Mixed Johnson Mtn 3 High Little Muddy 4 High Johnson Mtn 4 High Little Muddy 5 High Johnson Mtn 5 High Bear Creek 1 Mixed Beaver Creek 1 NFI Bear Creek 2 NFI Beaver Creek 2 NFI Bear Creek 3 High Beaver Creek 3 Mixed Bear Creek 4 NFI Beaver Creek 4 Mixed Bear Creek 5 NFI Beaver Creek 5 NFI Table 3. Historical fire severity designation for each plot. (NFI = not fire initiated or fire interval >250 years; ND = not dateable.) 20

23 have a third designation, not fire initiated or NFI (explained in further detail in Appendix E), which means that there is no evidence of fire in the stand at all. Fire may have played a role in such a stand, but it would have occurred so long ago that no evidence remains. Overall, we found that 45% of our plots were historically high-severity, 25% were mixed-severity, 25% were not fire-initiated (or time since last fire was so long that no fire evidence remained), and 5% were not determinable (Figure 8). All but one of the high-severity plots was located on the eastern side of the valley. (See Appendix F for a map of the plots with severity designations noted.) Two of the eastern sites, Johnson Mountain and Bull Elk Pond, were 100% high-severity, but while all the plots in the Bull Elk Pond site appear to date from the 1879 fires (i.e., all the aspen ND 5% NFI 25% Mixed 25% High 45% Figure 8. Proportion of plots (out of 40) at each historical severity designation. post-date 1879), the Johnson Mountain plots do not appear to have all burned in the same year (i.e., one plot burned at high-severity in 1879, but this fire did not occupy the entire site other plots burned at high-severity in earlier years). Taken together with the other three eastern sites (1, 2 and 4 highseverity plots per site), we can infer that these high-severity fires varied in their size/extent, not surprising given the heterogeneous topography of the valley. The number of plots showing evidence of mixed severity within a single plot also confirms our view that patches of low severity and high-severity fire are sometimes closely intermingled, and that intervals between successive high-severity fires at a point on the ground may vary widely. Just how large were the high-severity patches in these historical fires? The data that we collected are not sufficient to answer this question definitively, but we can make some preliminary interpretations. Recall that each site consists of a central plot plus four additional plots arranged to form a square that is 500 m on a side. Because all five plots in the Bull Elk Pond site burned at high severity in 1879, we can infer tentatively that this patch of high-severity fire was approximately the size of the site, i.e., 500 m x 500 m, for an area of 250,000 m 2 (= 25 ha = 60 acres). The actual size of the patch might be larger or smaller than 25 ha, since the fire may have continued burning at high severity for a considerable distance outside the area of our site or it may have burned at lower severity in places within the interior of the site. Nevertheless, only this one of our eight randomly located sites had evidence of high severity fire in the same year in all five plots; all of the other sites recorded 21

24 a mix of high severity, mixed severity, and non-fire initiated stands among the five plots. This pattern suggests that high severity patches as large as an entire site, i.e., on the order of 25 ha, were the exception rather than the rule; most patches apparently were smaller. We emphasize that this analysis is tentative, however. A new study with a different kind of sampling design would be needed to more rigorously document the the patch structure of historical fires. VI. DISCUSSION 1. Management Recommendations Many previous fire regime studies, particularly in ponderosa pine forest types, focused primarily on structurally-centered restoration practices in their management recommendations. In other words, dendroecological methods were used to ascertain the historical density, age structure, and species composition of the stands in question, and management techniques were geared toward achieving those historical standards. Various combinations of mechanical thinning and prescribed fire were then used by managers to meet these goals. Since that time, ecologists and managers have increasingly come to recognize that changing climate conditions dictate a more flexible restoration strategy, one based more on process than on structure (e.g., Falk 2006, Harris and others 2006, Millar and others 2007). More theoretical approaches include reconceptualizing ecosystems in economic terms, quantifying ecosystem goods and services and the processes that yield them (Harris and others 2006, de Groot and others 2002). In all cases, those interested in processcentered restoration (Falk 2006) recognize that many facets of future climate change are unknowable in the present, and thus require a management viewpoint that can adapt to changing conditions. The word toolbox often emerges in these papers, implying that priorities and goals will change over time, and that a variety of approaches will be implemented to achieve them. All of this points toward a management strategy based less on defined targets and outcomes, one wherein success may be less well-defined. Banded Peak Ranch management has already expressed an interest in focusing more on process than on structure we agree with this approach, as well as with the necessity of employing a toolbox of (sometimes conflicting) methods. In particular, we applaud management efforts to obtain burnover agreements with federal managers of adjacent properties. We also encourage management to follow the progress of Rocky Mountain National Park fire managers, who are hoping to experiment with prescribed crown burning over the next few seasons. Although the ecological situation there is quite different (i.e., masses of beetle-killed lodgepole pine trees that pose a hazard to park users), the outcomes in regard to policy (local, state, federal), public perception, potential escapes, etc. may prove useful in Banded Peak Ranch s planning for wildland fire use in the future. The integration of science and management is a complex task, with the means and ends of each often coming into conflict. To address the variety of questions posed by this 22