Temporal and Spatial Variability of Fire Occurrence in Western Oregon, A.D. 1200 – 2000

September, 2002

Evelyn L. Berkley
Space Imaging
421 SW Sixth Ave., Suite 850
Portland, OR, 97204, USA

Cathy Whitlock
Department of Geography
University of Oregon
Eugene, OR, 97403

Patrick J. Bartlein
Department of Geography
University of Oregon
Eugene, OR 97403

Frederick J. Swanson
USDA Forest Service
Pacific Northwest Research Station
Corvallis, OR 97331

Keywords: fire history; dendrochronology; spatial and temporal variability; climate conditions; vegetation conditions; Oregon; Pacific Northwest; synchroneity; time-series.

Link to web presentation of MS Thesis, December, 2000

Understanding past fire regimes in the Pacific Northwest enables better interpretation of effects of fire on forest structure and composition, as well as better assessment of management practices. Possible synchroneity in historic fire occurrence in western Oregon and concern about the representativeness of the available data motivated a synthesis of eight dendrochronological studies from western Oregon. Time-series images of past patterns were used to explore the spatial and temporal variability of fire occurrence for the period A.D. 1200 to present. Fire events were most widespread in the 1800s, particularly between 1850 and 1875, and fire events were widespread, but less numerous, in the 1500s. Many fire events occurred during the 1600s and 1700s, but they were localized and asynchronous. Study sites spanned the range of western Oregon climate and vegetation conditions with the exception of areas of very high and low precipitation.

Knowing that natural disturbance is a vital process in many native ecosystems obligates us to investigate its role in present forests and understand how and why disturbance regimes have changed over time. Historically, the predominant agent of disturbance in the Pacific Northwest was wildfire (Agee 1993). Fire frequency and severity have strongly influenced the structure and composition of forests in the region. Species dominance and watershed and ecosystem processes, such as nitrogen fixation and nutrient cycling, are also strongly influenced by fire (Agee 1993). Since the early 1900s, fire suppression has been an objective of forest management policy across the U.S. (Pyne 1982). Despite efforts to control fire, previous records of area burned in Oregon and the U.S. have been broken repeatedly during the past decade (Northwest Interagency Coordination Center 2000, Oregon Department of Forestry 2000, National Interagency Fire Center 2000), raising the question of whether or not the current incidence of fire has a precedent prior to Euro-American settlement.

Collecting and analyzing information on past fire events, including their size and frequency, is one way to address such questions (Swetnam et al., 1999). Pre-settlement fire history information primarily comes from fire event reconstructions based on dendrochronology and stand-aging methods. In complex fire regimes characteristic of the Pacific Northwest, dendrochronology is used to date fire-scars in tree ring patterns exposed on tree stumps, wedges, or cores and to determine establishment dates of early seral species. Fire-scarred trees provide information on non-lethal fire events using a combination of scar and tree establishment dates. Both methods may become less certain farther back in time as fewer living trees are of sufficient age to provide a record. Rules for interpreting a fire event differ somewhat among studies depending on the relative frequency of establishment and scar dates, as well as other factors.

The combination of wet winters and warm, dry summers characteristic of the Pacific Northwest results in low- to moderate-severity, high-frequency fire regimes in the wetter areas, and moderate- to high-severity, low-frequency fire regimes in the drier areas (Agee 1993). Weisberg and Swanson (in press) synthesized dendrochronological fire-history data of the last 600 years from ten fire-history studies in western Oregon and Washington. The percentage of each study area that burned during discrete time intervals was used to identify periods of widespread burning. Periods of extensive fires were reported from the A.D. 1400s to ca. 1650 and from ca. A.D. 1801 to ca. 1925. The first period ended at different times for different study areas, but a distinctive limited-extent fire interval between the two periods existed for all study areas. The patterns suggest strong regional synchroneity in the occurrence of large fires in western Oregon and Washington.

Although the synthesis presented by Weisberg and Swanson (in press) is an important step toward illuminating regional fire patterns in the Pacific Northwest, it raises additional questions about the spatial character of the data. For example, fires may have occurred at most sites within a study area or only at a few sites. Concurrent fires across the region may have been confined to sites with similar climate and vegetation conditions or dispersed among a range of site types. Also, the ten fire-history studies may not adequately represent the environmental diversity of the study region, thus giving a distorted picture of past fire regimes for the area. In this study, we address two questions: How well do existing fire-history study sites sample the range of climate conditions and vegetation types in western Oregon? How have the spatial and temporal patterns of fire occurrence changed through time in western Oregon?

Focusing on eight dendrochronological studies from western Oregon, six of which were used in Weisberg and Swanson (in press), we first examined the location of reconstructed fire events within the context of climate and vegetation. A "fire event" can be one fire or it can be multiple fires that occurred close together in time (the temporal precision of dendrochronologic methods is rarely high enough to permit a distinction). We demonstrated with visualization techniques where sample sites were located and where fires occurred not only in geographic space, but also in climate and vegetation space. Second, we mapped and summarized spatial and temporal patterns of fire occurrence among all of the study areas. The studies that were selected were not originally developed as part of a unified project, but similarities in methodology and scale warranted their inclusion in the synthesis. We hope this effort will help researchers more efficiently target study areas that could elucidate regional fire patterns.

Fire-history data from eight dendrochronological studies (Fig. 1) were compiled for this synthesis and joined with environmental data for the same locations.

Figure 1: Location of study areas.

The particular studies were chosen based on their large spatial extent, long period of record, and relatively high sampling density (Table 1). The study areas ranged from 66 to 1375 km² in area and thus were large enough to reveal patterns at a landscape scale. The earliest fire record was A.D. 1200 in a few study areas, but most records began in the 1500s. The sampling density of the studies ranged from 0.1 to 3.6 sites/km². Most of the data used in the fire-history reconstructions were collected from tree stumps, but few tree-ring records were cross-dated because the objectives of the studies emphasized large spatial extent and high sampling density rather than temporal precision. Counting of tree rings on stumps results in errors in estimation of tree and scar age (Weisberg and Swanson 2001).

Table 1: Summary of study sampling designs and environmental characteristics of study areas.

The northernmost study area is the Bull Run watershed, the primary source of water for the Portland metropolitan area (Agee and Krusemark 2001). Most of the watershed lies within the Tsuga heterophylla Zone, and western hemlock (Tsuga heterophylla) and Douglas-fir (Pseudotsuga menziesii) are the dominant species. Higher elevation sites are in the Abies amabilis Zone, which is populated by species such as western hemlock and Pacific silver fir (Abies amabilis) in undisturbed areas. To reconstruct past fires, aerial photos were first used to identify distinct cohorts of trees, which were presumed to have established following fire. Initiation dates for each cohort were then estimated by counting rings on wedges, stumps, or increment cores.

The central Coast Range study has the largest spatial extent of all the studies, 1375 km² (Impara 1997). Most of the vegetation is in the Tsuga heterophylla Zone, and Douglas-fir is typically the dominant species, except along the coast where Sitka spruce (Picea sitchensis) is dominant. Dendrochronological data were collected throughout the study area using a scale-hierarchical design that stratified sampling by precipitation zone, land-type association, aspect, and hillslope position.

The Coburg Hills are located in the foothills of the Cascade Range at the eastern edge of the Willamette Valley (Weisberg 1995). The study area ranges from 100 to 930 m in elevation, and is within the Tsuga heterophylla Zone. Douglas-fir is the dominant tree species, although western hemlock, western redcedar (Thuja plicata), and incense cedar (Calocedrus decurrens) are also common. Big leaf maple (Acer macrophyllum), giant chinkapin (Chrysolepsis chrysophylla), red alder (Alnus rubra), and Oregon white oak (Quercus garryana) are common hardwoods. Sampling was conducted in clearcuts and sites were stratified over three elevation zones, resulting in a sampling density of 0.1 sites/km² (Weisberg 1995). Within each site, transects were established at different hillslope positions, and fire-scar ages and tree-origin dates for early-successional tree species were obtained. In addition, old tree stumps with plentiful fire-scar evidence were sampled opportunistically.

The Bear-Marten Watershed is located in the central western Cascades and surrounds a segment of the McKenzie River (Weisberg 1997). The Bear-Marten watershed is in the Tsuga heterophylla Zone. Fire-scar ages and tree-origin dates were determined by tree-ring counts in the field. Eleven of 63 sites—located in the northeastern, eastern, and southwestern sections of the watershed—were sampled intensively, and the remaining sites were sampled with a combination of systematic and opportunistic approaches. Twenty cross-sections at five sites were also removed and cross-dated.

The Blue River study area, northeast of Bear-Marten watershed, encompasses several small watersheds that feed into the McKenzie River. Most of the study area is within the Willamette National Forest, with the H.J. Andrews Experimental Forest occupying about 14% of the total area (Weisberg 1998). The Blue River study area is characterized by steep and dissected topography. Areas up to about 1000 m in elevation are in the Tsuga heterophylla Zone, and higher elevations are generally in the Abies amabilis Zone. Douglas-fir is the dominant species in the lower zone, whereas Pacific silver fir, noble fir (Abies procera), and western hemlock, in addition to Douglas-fir, are dominant in the upper zone. The highest elevations are in the Tsuga mertensiana Zone where mountain hemlock (Tsuga mertensiana) is common. A randomly selected clearcut from each legal section, averaging 2.79 km² in area, was chosen for sampling. Of 137 sampled clearcuts, 90 were sampled with a combination of systematic and opportunistic approaches, and 47 were sampled only opportunistically due to rough terrain.

The Augusta Creek watershed is located just west of Chucksney Mountain and drains into the South Fork of the McKenzie River in the western Cascades. The small study area is composed of several distinct landforms, including areas of highly dissected topography and cirque basins, as well as wide, glacially-carved valley floor (Kertis et al. 1995, Cissel et al. 1998). About three-quarters of the sites were in the Tsuga heterophylla Zone, and the remainder were in the Abies amabilis and Tsuga mertensiana zones. About 78% of the sample sites are in clearcuts, where stumps were prepared and counted. The other sites were forested and required increment cores to be taken from live trees. Sampling frequency was increased in areas of varied topography and where scar histories appeared complex.

The Cascade Crest study area is located in the upper elevations (1100 – 2000 m) of the central Cascade Range (Kertis and Huff 1997). Most sites were within the Three Sisters Wilderness Area, although some extended into the northern portion of Waldo Lake Wilderness Area. The study area includes portions of both the Abies amabilis and Tsuga mertensiana zones. Site locations were chosen based on a nested sampling scheme stratified by areas of similar vegetation characteristics and topography (Kertis and Huff 1997). Fire history was reconstructed using stand-age analysis based on tree-ring counts and aerial photos, as well as using other fire evidence, such as charred wood or soil.

Little River is the southernmost study area and occupies most of the Little River watershed, a tributary of the North Umqua River (Van Norman 1998). The terrain consists of mostly steep slopes and narrow valley floors. Little River is located in an ecotone between the Mixed Conifer and Tsuga heterophylla zones. Douglas-fir dominates at low to mid-elevations, whereas western hemlock is common at mid- to high elevations. Several other conifer and hardwood species are found at low to mid-elevations, depending on the amount of moisture available. At dry low to mid-elevations are incense cedar, sugar pine (Pinus lambertiana), and grand fir (Abies grandis). At moist low to mid-elevations, western redcedar and Pacific yew (Taxus brevifolia) occur. One or more clearcuts were sampled from each cell within a 2-km² grid covering the study area. Where terrain was fairly homogeneous, a transect was randomly placed at each site. All stumps at the beginning of each transect were sampled, and additional stumps were sampled further along and outside the transect to verify fire dates.

In an ideal study design, one would try to explain the results of the second research objective—how the spatial and temporal patterns of fire occurrence have changed through time in western Oregon—by comparing the fire-history records with independent records of past climate for the same region. Because paleoclimate data are not available for the entire study region (Graumlich 1987), it was necessary to use modern climate as a basis for assessing geographic patterns. Although the climate has changed over the last 800 years, relative differences between study areas have probably changed little (Mock 1996), therefore we thought it was reasonable to use modern data to investigate whether fire events tend to be stratified by climate or vegetation and how well the study areas represent the climate field of western Oregon.

The variables used in the climate analysis are based on western Oregon weather station data for the period A.D. 1951 – 1980. Temperature, precipitation, and percent possible sunshine data were interpolated onto a 2.5-min grid of western Oregon using a procedure that considers the influence of elevation (Thompson et al. 1999). These values were also interpolated for the fire-history sites. Bioclimatic variables that influence the distribution of plants (Prentice et al. 1992, Sykes et al. 1996) were also derived from the climate data. The two variables used were soil moisture, indicated by actual evapotranspiration divided by potential evapotranspiration, and length of growing season, indicated by growing degree days above 5 °C.

Two vegetation datasets were used in this study: current vegetation, as defined by the Gap Analysis 2 (GAP2) land cover map, and potential natural vegetation, as determined by a U.S.D.A. Forest Service model. The GAP2 data (2000) were obtained from the Oregon Geospatial Data Clearinghouse, and documentation was provided by the Oregon Natural Heritage Program (Kiilsgaard 1999). Potential natural vegetation (PNV) data generated by the U.S.D.A. Forest Service represents the vegetation that would exist if current forests reached a climax community state in the absence of catastrophic disturbance (Henderson 2001). The main limitation of the PNV dataset was that it did not cover the entire study region (the extent of western Oregon encompassing the study areas) at the time of the analysis.

The initial phase of this project involved compiling the fire-history data for all of the sites within each study area into a single fire-history database. The second phase consisted of adding information about the environment of each site to the database. Data were added for coordinate location, elevation, climate conditions, and current and potential vegetation type. In the third phase, visualization and analysis techniques were used to explore the spatial and temporal patterns of fire occurrence in the study region. Patterns were studied in geographic space (i.e., on a map of western Oregon), as well in the environmental space of western Oregon (i.e., on diagrams showing climate and vegetation gradients of western Oregon).

The fire-history data were obtained from the researchers directly or from the Forest Science Data Bank (2000), a partnership project between the Department of Forest Science, Oregon State University, and the U.S.D.A. Forest Service Pacific Northwest Research Station in Corvallis, Oregon. For each study area, an ArcInfo or ArcView file containing the point locations of the individual sites was obtained. The fire-event data for each study were available in a spreadsheet or, in a few cases, as part of the GIS point coverage of locations. When necessary, data were reformatted to create one row for each site and one column for each year in which fire events occurred. The resulting fire chronology contained data indicating "no fire event" or "fire event" for every site in every year included in the chronology. If researchers had distinguished between "no fire event" (e.g., no fire-scars corresponded to a given date) and "no data" (e.g., no trees existed that were old enough to have recorded fire at a given date) this information was preserved.

The fire-event chronologies were then joined to the corresponding GIS point coverages (ArcInfo version 8.0.2). The final result was a unique GIS coverage of sites for each study area, including a map of site locations and a corresponding point attribute table containing the years in which fire events occurred at each site. Because half of the studies did not include estimates of the fire severity of individual fire events, fire severity was not used in the analysis.

Topographic information was extracted from U.S. Geological Survey and U.S.D.A. Forest Service digital elevation models (DEMs) (1997) following conversion to ArcInfo Grid format. A public domain AML (a program written in Arc Macro Language) was used to extract the elevation, slope, and aspect for sites within each study area from the gridded DEMs. The AML added these values to the point attribute table in the corresponding GIS coverage. A shaded-relief map of western Oregon was also generated to provide a locational context for the fire-history data. The data for the shaded relief were obtained from GTOPO30 (1996), a global DEM with approximately 1-km resolution that is available from the U.S. Geological Survey, and converted to an index of gray shades.

Two climate datasets were prepared for use with the fire-history database. One contained the latitude and longitude values and climate conditions of the sites, and the other contained the same parameters for all of the points in a 2.5-min (~4.6-km) grid covering western Oregon.

The vegetation data were prepared in a manner similar to that of the climate data. The GAP2 vegetation data were exported from ArcInfo and used to generate two separate datasets, one containing current vegetation types for the sites and the other containing current vegetation types for all of western Oregon. A few classes were combined (e.g., different wetland types were merged into one class called "wetland"). Two subsets of data were also extracted from the potential natural vegetation grids. The first contained the PNV Plant Association Groups for the sites, and the second contained the PNV Plant Association Groups for points on the grid of western Oregon. Because no modeled data were available for southwestern Oregon, the sites and grid points that fell within that area received a "no data" value. The PNV Plant Association Groups (e.g., Douglas-fir/vine maple/western fescue or Douglas-fir/common snowberry) were then aggregated according to potential dominant canopy vegetation (e.g., Douglas-fir zone) in order to facilitate the analysis. (Each PNV vegetation zone, which differs from the more generalized Franklin and Dyrness (1973) vegetation zones, is characterized by one dominant tree species for which the zone is named, but within a zone, a variety of combinations of understory vegetation can exist.) The potential and existing vegetation data were then combined with the two climate datasets.

To address the first research question, regarding the representativeness of the sites, scatter diagrams were generated to compare the climate and vegetation of the sites to the entire range of climate and vegetation conditions of western Oregon.

The second question dealt specifically with fire occurrence and was addressed by creating a time series of images that reveals the temporal and spatial components of the data. This required three steps: First, an appropriate time interval was chosen for binning the fire data. Although the fire events in the fire-history database were associated with individual years, the temporal resolution reported by the researchers varied. A study of dating errors in Douglas-fir forests of the central western Cascades demonstrated that about 75% of fire dates estimated by field counts of fire-scarred tree rings are likely to be within ten years of the date determined by cross dating, and about 87% are likely to be within 20 years of the actual date (Weisberg and Swanson 2001). Thus, 25 years, which was also the lowest estimated temporal resolution reported by any of the studies, was selected as the binning interval. The fire-history dataset was summarized by 25-year intervals at a 25-year step, beginning with A.D.1200 and ending with A.D. 1975. Trials with other discrete and overlapping time intervals, ranging from one to 30 years, verified the robustness of 25 years as a basis for analysis.

The second step was to create maps of fire occurrence in geographic space for each 25-year interval. Interactive Data Language (IDL version 5.2) was chosen as the programming language and visualization environment in which to create the maps because it permits a high level of control over the output and has built-in mapping functionality. Fire-history sites on the time-series maps were differentiated as having a fire event in the 25-year time interval, as "no data" (i.e., sampled trees were not alive during the given 25-year period), or as "no fire" (i.e., trees were alive during the specified time period, but did not bear any evidence of fire during that period).

Relatively few stands over 500 years old remain, resulting in only a partial fire record at most sites for the earliest centuries; therefore many symbols indicate "no data" at the beginning of the time series of images. Unfortunately, half of the studies (Bull Run watershed, Augusta Creek, Cascade Crest, and Little River) did not distinguish in their fire history reconstructions between "no data" and "no fire". Thus, if all the non-burning sites for those areas were shown as "no fire" symbols, the resulting map might under-represent the actual number of sites that burned. Although this problem is less critical in recent centuries when most sites had trees of sufficient age, it is a concern for earlier centuries. In order to minimize misrepresentation for the Bull Run, Augusta Creek, Cascade Crest, and Little River study areas, all sites were designated "no data" until the first fire event was recorded anywhere within one of the study areas. During this time interval, all of the "no data" symbols in the study area, except the one(s) that experienced fire, change to "no fire" symbols, indicating that data exist, but no fire occurred. Some under-representation of fire undoubtedly remains, but this technique at least reduces over-interpretation of the record when no data exist.

Third, the time series of images was created by making a separate layout containing a geographic map and set of climate and vegetation scatter diagrams for each of the 32 time intervals. To summarize the data presented in the time-series, density plots (Venebles and Ripley, 1999) were created for each study area, showing the variations in the frequency of fire events over time. Density plots provide a smoothed depiction of the relative frequency of events (number per unit time) that are not as sensitive to the choice of parameters as, for example, histograms are to the choice of the bin width. Only data for the period A.D. 1500 – 2000 were used because the data available prior to A.D. 1500 are so sparse.

Scatter diagrams allow comparison of the climate and vegetation range of the sites with the climate and vegetation range of western Oregon (Fig. 2a-c). The study sites represent the range of elevation and cover most of the climate space that exists in western Oregon. Extremely wet and extremely dry climate types are not well sampled.

Figure 2a

Figure 2b

Figure 2c

Figures 2a-c: Correspondence between geographic space and data space of study sites and western Oregon. Some symbols are obscured in the scatterplots because they are overlapped by others.

The vegetation data are discrete variables, rather than continuous variables like the climate data, and therefore appear in distinct columns instead of clouds of points. The two vegetation classification systems present the vegetation classes roughly in order of increasing elevation. The current vegetation diagrams (GAP2 Vegetation vs. Elevation) indicate that 25 land cover classes are present in western Oregon (Table 2). Vegetation types most extensively sampled by the sites are Douglas-fir/western hemlock/western redcedar forest, forest/grassland mosaic, mixed conifer/mixed deciduous forest, true fir/hemlock montane forest, and subalpine fir/lodgepole pine montane conifer. Sitka spruce/western hemlock maritime forest and mountain hemlock montane forest, although less common in western Oregon than the other GAP2 types, are also well sampled by the sites. Current vegetation types that are most under-sampled by the sites are wetland, agriculture, Oregon white oak forest, mixed deciduous forest, ponderosa pine/white oak forest and woodland, lodgepole pine forest and woodland, and ponderosa pine dominant mixed conifer forest. Areas of western Oregon for which there were no GAP2 data are designated by "N/A"

Table 2: Vegetation key. Explanation of GAP2 and potential natural vegetation codes used in Figure 4.

The potential vegetation diagrams (Potential Vegetation vs. Elevation) reveal 15 PNV zones in western Oregon, but the dataset only extends to the north edge of the Little River study area and to the crest of the Cascade Range. Areas with no data, including all sites in Little River and a few sites in the Cascade Crest study area, are labeled "N/A". The remaining sites extensively sample four of the PNV zones in western Oregon: western hemlock, Pacific silver fir, Douglas-fir, and mountain hemlock. The Sitka spruce zone is also relatively well sampled. Poorly sampled PNV zones are the grand fir, mountain hemlock-subalpine fir, and lower-elevation white fir zones. Mountain hemlock-subalpine fir, a mixed zone, is different from the distinct mountain hemlock and subalpine fir zones, and occupies the highest elevations of all of the PNV zones, except the Alpine zone.

The binned fire events display the temporal distribution of the site records from each study area (Fig. 3). Given the coarse temporal resolution of a 25-year interval, the symbols may signify more than a single fire event. The largest cluster of fire events occurs between ca. A.D. 1775 and 1900, and the second largest cluster is between ca. A.D. 1500 and 1625. Fewer sites burned in the years between these two periods. Individual exceptions are sites in the Bull Run, Coast Range, Coburg Hills, Augusta Creek, and Cascade Crest study areas, all of which also experienced brief, asynchronous periods of fire occurrence between ca. A.D. 1625 and ca. 1800.

Figure 3: Fire-event data binned at 25-year intervals with a 25-year step. Black circles represent sites that had one or more fire events within each 25-year interval. These data were used in constructing the time-series maps and scatter diagrams in Figure 4.

The time-series of images shows which sites burned during each time interval (Fig. 4); the actual number of fire events that occurred at each site is not revealed because the fire data presented in the time-series are binned. Data are sparse in the early part of the record from A.D. 1200 – 1300. One known site burned in the Cascade Crest study area between A.D. 1200 and 1225, and five burned in the Bull Run watershed in the northern portion of the Cascades between A.D. 1225 and 1250. Most of the other studies did not contain trees of sufficient longevity and number, and thus are shown as having no data. There was no evidence of fire in any study area from A.D. 1300 to 1400, but few sites had trees old enough to record fire during this time. The most striking pattern during the A.D. 1400s is that 97% of the sites in the Bull Run watershed burned between A.D. 1475 and 1500, a phenomenon that was attributed to a single fire event that swept the area in A.D. 1493 (Agee and Krusemark 2001). Also, the period A.D. 1475 – 1500 marked the first interpreted fire events in the Blue River and the Coast Range study areas. The burned Blue River sites were at relatively high elevations and on both sides of the Blue River valley, and the burned Coast Range sites were scattered in the eastern, interior half of the study area. A small cluster of sites in the southern part of the Little River study area in southwestern Oregon experienced fire events at different times throughout the last three-quarters of the 15th century.

Figure 4: Time series animation of fire occurrence in geographic, climate, and vegetation space. See Table 2 for vegetation key. The Macromedia Flash plug-in is needed to view the animation and can be downloaded for free at the following URL: http://sdc.shockwave.com/shockwave/download/frameset.fhtml?P1_Prod_Version=ShockwaveFlash

The A.D. 1500s are marked by a sharp increase in the number of burned sites, reaching a maximum of 15% of all sites, followed by a decline to 9% and then 11% by the end of the century. The first recorded fire events in the Augusta Creek study area occurred between A.D. 1500 and 1525 and were widely distributed spatially. Although data are still unavailable for the beginning of this century in the Coast Range, the Coburg Hills, and the Bear-Marten watershed, all studies have at least some data for the period A.D. 1525 – 1550 and onward. During these years, several sites burned in both the western and eastern sides of the Coast Range study area.

Between A.D. 1550 and 1575, fire events in the Blue River study area shifted from mostly high-elevation eastern sites to low-elevation western sites. Sites burned in the highest elevations of the Cascade Crest study area, as did sites in the eastern portion of the Little River study area. Between A.D. 1575 and 1600, the spatial distribution of burned sites was very different than in the rest of the 16th century. Most sites in the Blue River, Augusta Creek, and Cascade Crest study areas experienced fire events during this period. Within these areas, burned sites were widely distributed. Only a few sites, on the east side, burned in the Coast Range. Two sites burned in the northern part of the Little River study area, and no sites burned in the Bull Run watershed, the Coburg Hills, or the Bear-Marten watershed. For study areas that did burn, fire events generally shifted to higher-elevation, wetter sites during the last quarter-century.

A general shift in fire occurrence to lower-elevation, dry sites took place at the beginning of the A.D. 1600s. The total percentage of sites burned was consistently around 7% for each 25-year interval in this century. No sites burned in the Coast Range from A.D. 1600 to 1625, and fires during the last three-quarters of the 17th century and in the 150 years thereafter were limited to the eastern margin of the study area. About 25% of sites in the Blue River study area burned from A.D. 1600 to 1625, but this percentage dropped during the rest of the century. More sites (18%) recorded fires in the Little River study area than during any previous period. A few sites burned in the Bull Run watershed, the eastern part of the Bear-Marten watershed, and the northern and southern sections of the Cascade Crest study area.

The period from A.D. 1625 to 1650 featured many fire events in the Coburg Hills, as well as clustered events in the central Cascade Range, including the southeastern part of Bear-Marten watershed, the Blue River valley, and the northern part of Augusta Creek. During the period A.D. 1650 – 1675, fires were not registered in the Coburg Hills nor in most of the central Cascade Range study areas, but the number of fire events increased in the Cascade Crest study area and in the southeastern section of the Bull Run watershed. Some Cascade Crest fire events were in the wettest sites of the study area, which had not burned in prior centuries. The percentage of total sites that burned from A.D. 1675 – 1700 was similar to that from A.D. 1625 – 1650, yet the distribution was different. Fires in the Coburg Hills during this last quarter-century were located along the western margin of the study area, closer to the Willamette Valley, and those in the Blue River study area were mostly in the northern portion, an area characterized by relatively gradual slopes. Sites that burned in the Bull Run and Cascade Crest study areas were confined to the drier, lower-elevation southwestern corner and southern half, respectively. Sites that burned in the Little River study area were all in the upper reaches of the watershed.

The A.D. 1700s resembled the A.D. 1500s in terms of percentage of total sites burned, however, the spatial distribution across all studies was more uneven. As in the A.D. 1600s, the sites that experienced fire in the Coast Range were limited to the eastern portion. From A.D. 1700 to 1725, no sites burned in the Augusta Creek study area. A few sites in the Coburg Hills and the northern part of the Bear-Marten watershed, several sites in the Cascade Crest study area (including a few high-elevation, wet sites), and many in the Blue River and Little River study areas burned. From A.D. 1725 to 1750, some sites in the Coast Range burned, although fires were still confined to the east. Fire events shifted to the southwest in the Coburg Hills and parts of the central Cascade Range (Bear-Marten and Blue River study areas). Several sites in the Augusta Creek and Little River study areas and a few lower-elevation sites in the Cascade Crest study area registered fire. The period A.D. 1750 – 1775 was much like the previous 25 years, with a slight increase in the number of sites burned in most study areas. Fire events in the Blue River study area shifted from central and southwestern sites to higher-elevation, northeastern sites. The percentage of affected sites in the Coburg Hills (63%) was higher than in any study area during any previous period. From A.D. 1775 to 1800, the number of fire events in all study areas decreased or stayed nearly the same except in the Augusta Creek area, where the percentage of burned sites increased to 35. The Bull Run watershed did not experience any fire events in the A.D. 1700s.

Fire patterns changed across the study region during the A.D. 1800s. From A.D. 1825 to 1850, the total percentage of burned sites increased from 14 to 22%. The most dramatic increase in fire activity came between A.D. 1850 and 1875 when 32% of the sites burned. For the first time in the record, the spatial distribution of fires was relatively even in all study areas during the same 25-year period. Even the wettest, Sitka spruce-dominated sites near the coast were affected. Between A.D. 1875 and 1900, the percentage of burned sites decreased to 18%, although it was still high relative to other periods. The decrease was most pronounced in the Coast Range where the percentage dropped from 69 to 24%.

A dramatic decline in the number of sample sites recording fire occurred in the A.D. 1900s. During the first quarter-century, western sites in the Coast Range recorded no fires and those in the Blue River study area were concentrated in the dryer, more accessible west and south. Fire events continued to occur near the study area boundaries in the Bull Run, Augusta Creek, Cascade Crest, and Little River study areas. The percentage of burned sites actually increased from the previous period by 20% in the Bear-Marten watershed. The most striking decline began in the period A.D. 1925 – 1950, when the percentage of total burned sites dropped from 17 to 5%. Fire events during this time tended to be at low elevations in the Coast Range, the Coburg Hills, and the central Cascade Range (the Bear-Marten watershed and southern sites in the Blue River study area). Single sites burned in the Bull Run watershed and in the Cascade Crest study area. The decline continued from A.D. 1950 to 1975, when three low-elevation sites burned in the eastern and western portions of the Bull Run watershed, a few sites burned in the western half of the Coast Range, a few burned in the study areas in the central Cascade Range, and several scattered sites burned in the Coburg Hills. By the period A.D. 1975 – 2000, the only sites that experienced fire events were three in the Little River valley, and one in the McKenzie River valley at the southern edge of the Blue River study area.

The density plots of the unbinned fire event data from A.D. 1500 to 2000 present a useful summary of the temporal patterns of fire occurrence (Fig. 5). The temporal truncation at the beginning of the record affected the density curve for the Bull Run watershed because nearly all sites in that area burned in A.D. 1475 – 1500. If the density curve began at A.D. 1475, the highest peak for the Bull Run watershed would be from A.D. 1475 to 1500 and the other peaks would be much lower. Most of the major peaks for the different study areas occurred between A.D. 1850 and 1900, with the highest density of fire events occurring between A.D. 1850 and 1875. Half of the study areas (the Coast Range and the three in the central Cascade Range) show pronounced peaks in fire event density in the A.D. 1500s. Some study areas feature secondary peaks in the A.D. 1600s (the Bull Run watershed) or A.D. 1700s (the Coburg Hills, Cascade Crest, and Little River study areas).

Figure 5: Density plots showing the variations in the relative frequency of fires (number of fires per unit time). The area under a section of the curve reflects the local proportion, or density, of fire events during the given time interval relative to the total number of fire events recorded in the study area from A.D. 1500 to 2000. Data are smoothed using a 25-year window and 25-year step.

This investigation emphasizes the spatial component of past fire events and assesses how representative the sites are of western Oregon environments. This approach and that of Weisberg and Swanson (in press) both reveal clearly a period of elevated fire activity in the late 1800s. However, the time-series maps (Fig. 4) imply that the period of synchroneity was somewhat more focused (i.e., ca. A.D. 1825 – ca. 1900, rather than ca. A.D. 1800 – ca. 1925). Both approaches show increased fire activity in the A.D. 1500s followed by a period of reduced fire activity in the A.D. 1600s and A.D. 1700s.

It is interesting, though not entirely surprising, that western sites in the Coast Range burned only in the A.D. 1500s and 1800s. Fire-return intervals in Sitka spruce forests along the Pacific coast may be on the order of hundreds of years (Agee 1993). Thus, these forests tend to burn only under extreme fire conditions, often a synergy of drought, strong east winds, and ignition in the Willamette Valley or eastern side of the Coast Range (Impara 1997). These wet sites may also be valuable in fire-history reconstructions because they can show changes in fire frequency and severity in response to extreme drought.

Only five intervals feature fire events in two or more of the wettest sites from either the Bull Run watershed, the Coast Range, or the Cascade Crest: A.D. 1475 – 1500 (Bull Run), A.D. 1550 – 1575 (Coast Range and Cascade Crest), A.D. 1650 – 1675 (Cascade Crest and Bull Run), A.D. 1700 – 1725 (Cascade Crest), and A.D. 1850 – 1875 (Coast Range). A.D. 1475 – 1500 is the only period in which all sites in a study area (Bull Run) burned, possibly a consequence of severe drought in conjunction with high fuel loads and a fortuitous combination of ignition and spread conditions. Without data from other sites during that time, it is difficult to speculate on other scenarios. The second period, A.D. 1550 – 1575, may have been characterized by dry conditions that led to widespread fires, whereas the third and fourth periods, A.D. 1650 – 1675 and A.D. 1700 – 1725, may have been characterized by more localized drought. The increase in fire activity in the Bull Run and Cascade Crest areas between A.D. 1650 and 1675 coincides with dendrochronological reconstructions of a brief warming period from A.D. 1650 to 1690 in Longmire, WA, further north in the Cascades (Graumlich and Brubaker 1986). The impact of this warming on fire activity was relatively insignificant compared with other periods, however, perhaps because its short duration did not allow sufficient fuels to accumulate in areas that had burned in the A.D. 1500s. Finally, the fifth period, A.D. 1850 – 1875, was marked by widespread fires like the first period, but the only extremely wet sites that burned were in the Coast Range. Absence of fire in the wettest Cascade Crest or Bull Run sites during this period suggests that either fuel load was low (although the wettest Cascade Crest sites had not burned for 100 – 200 years) or that burning was largely a result of human activities rather than climate. Burning of wet sites along the Oregon Coast may have been influenced by European settler ignitions or strong east winds. Other settler influences, such as sheep herding and road building, probably contributed to a non-uniform distribution of fire events in some study areas (Burke 1979).

This investigation highlights exceptions to the wholesale decrease in fire activity in the A.D. 1600s and 1700s depicted in Weisberg and Swanson (in press). During the mid-1600s, the number of burned sites actually increased from the previous interval in the Bull Run and Cascade Crest study areas. Fire activity in the Bull Run watershed is not shown for this time period in Weisberg and Swanson (in press) because cumulatively over 25 years, < 20% of the study area burned. Upper montane forests in the Cascade Range are usually dominated by Pacific silver fir, a species that is particularly vulnerable to fire because of its thin bark. Consequently, these forests are associated with infrequent high-severity events, often brought on by summer drought and/or strong east winds (Agee 1993).

A steady increase in fire incidence occurred throughout the A.D. 1700s in the percentage of sites that burned in the Coburg Hills, Augusta Creek, and Little River. The Cascade Crest study area and the Coburg Hills experienced a distinct, coincident peak in fire occurrence between ca. A.D. 1750 and ca. 1775 (Fig. 5). These increases occurred well in advance of European settlement, leaving Native American practices, climate, and/or local weather conditions as possible contributing factors. The Molala are known to have inhabited the lowlands of the western Cascade Range, and archaeological evidence also indicates that they occupied higher elevations during summer (Aikens 1993). Thus, some parts of the study areas experienced protracted use by native people (e.g., the Coburg Hills, the coastal strip of the Coast Range, and low elevation sites in both the eastern portion of the Coast Range and the western portion of the Little River study area). Other areas may have had quite limited visitation (e.g., the more remote parts of the Blue River, Augusta Creek, and Bull Run study areas). As Weisberg and Swanson (in press) suggest, higher resolution fire-history reconstructions would be helpful in discerning climate and human impacts during the late A.D. 1700s and early A.D. 1800s, but important aspects of early human influences will probably never be known.

Although the time series of images effectively communicates the spatial distribution of interpreted fire events from A.D. 1200 to present, the regional scale patterns over space and time are best summarized by the density plots (Fig. 5). Because the peaks are scaled to show the number of fire events for each interval as a proportion of the total events for the entire 500 year period in each study area, they emphasize periods of unusually high fire occurrence and de-emphasize periods of low fire occurrence. Coincident peaks suggest periods of increased fire occurrence on a regional scale. In comparing the peaks among studies, it is evident that there was regional synchroneity in the mid- to late A.D. 1800s and in the A.D. 1500s, although the total number of fire events was much lower in the earlier period. However, fewer trees were of sufficient age to provide a record in the A.D. 1500s, so fire events are likely underrepresented for this century. At smaller spatial scales and shorter timescales, fires were synchronous between some studies in the intervening years (e.g., the Bull Run watershed and the Cascade Crest study area from A.D. 1650 to 1675, and the Coburg Hills and the Cascade Crest study area from A.D. 1750 to 1775).

This mapping of existing dendrochronologic fire history information sets the stage for further analysis of temporal and spatial patterns of fire history in the region. This field of research is at a point well suited for geographically explicit analysis of influences of Native- and Euro-Americans on fire ignitions and suppression across the landscape. The timing and geographic patterns of these human influences were quite varied. To comprehend the variability of forest dynamics and fire regimes on longer timescales, long paleoecological records are needed. Recent research on Holocene fire and vegetation history in the Coast Range suggests that fire frequency has varied with climate change on millennial timescales, and current fire regimes may be a phenomenon only of the last 1000 years (Long et al. 1998). Further analysis of the temporal dimension of climate-fire interactions can take the form of extended paleoecological studies of the longer-term climate-fire context of the region nested within this dendrochronologic framework, higher resolution dating of fire events to test the time scale of apparent regional synchroneity, and dendroclimatology to enhance understanding of climate variation in western Oregon. These approaches to historical interpretation can be complemented with modeling (e.g., Wimberley et al. 2000). Reconstruction of regional vegetation dynamics requires such a multi-dimensional approach, given the strengths and weakness of any given analytical tact.

Several of the individual studies considered here were undertaken to gain understanding of the historic fire regime for use in local landscape management plans (Blue River (Cissel et al. 1999), Augusta Creek (Cissel et al. 1998), Bear-Marten, Little River). However, this regional synthesis suggests significant climate influence on fire occurrence at the inter-century time scale. This observation focuses attention and importance on the questions: What is the relevance of the past to management of the future, especially in the context of climate change? What are the ecological and social consequences of various types and degrees of departure of the managed landscape from the historical range of landscape conditions? Attempts to implement management plans based in part on historical disturbance regimes, in conjunction with lessons from management schemes based largely on principles of conservation biology and intensive plantation forestry, will provide useful information for addressing these questions in a changing climate.

This research was made possible by a Co-operative agreement with the USDA Forest Service (USFS PNW-9805122-1CA) and grants from National Science Foundation to the H.J. Andrews Experimental Forest Long-Term Ecological Research site, to Patrick J. Bartlein (ATM-9910638), and to Cathy Whitlock and Patrick J. Bartlein (ATM-0117160). GIS support was provided by the InfoGraphics Lab at the University of Oregon Geography Department.

Literature Cited

Agee, J. K., and F. Krusemark. 2001. Forest fire regime of the Bull Run watershed, Oregon. Northwest Science 75(3):292-306.

Agee, J. K. 1993. Fire ecology of Pacific Northwest forests. Island Press, Washington D.C.

Aikens, M. 1993. Archaeology of Oregon. U.S. Department of Interior, Bureau of Land Management. BLM/OR/WA/ST-93/16+8100.

ArcInfo version 8.0.2. 2000. Computer software. Environmental Systems Research Institute. Windows NT.

Burke, C.J. 1979. Historic Fires in the Central Western Cascades, Oregon. M.S. thesis, Oregon State University, Corvallis, Oregon.

Cissel, J.H., F.J. Swanson, and P.J. Weisberg. 1999. Landscape management using historical fire regimes: Blue River, Oregon. Ecological Applications 9(4):1217-1231.

Cissel, J.H., F.J. Swanson, G.E. Grant, D.H. Olson, S.V. Gregory, S.L. Garman, L.R. Ashkenas, M.G. Hunter, J.A. Kertis, J.H. Mayo, M.D. McSwain, S.G. Swetland, K.A. Swindle, and D.O. Wallin. 1998. A landscape plan based on historical fire regimes for a managed forest ecosystem: the Augusta Creek study. U.S.D.A. Forest Service General Technical Report PNW-GTR-422. Pacific Northwest Research Station, Portland, OR.

Digital Elevation Models. 1997. Dataset. Oregon Geospatial Data Clearinghouse. URL: ftp://ftp.sscgis.state.or.us/pub/data/baseline97/raws.

Forest Science Data Bank. 2000. H.J. Andrews Long-Term Ecological Research Program. URL: http://www.fsl.orst.edu/lter/data.cfm?topnav=8.

Franklin, J. F., and C.T. Dyrness. 1973. Natural vegetation of Oregon and Washington. U.S.D.A. Forest Service General Technical Report PNW-8.

GAP2 Vegetation Data. 2000. Dataset. Oregon Geospatial Data Clearinghouse. URL: ftp://ftp.sscgis.state.or.us/pub/data/statewide/k100.

GTOPO30. 1996. Global digital elevation dataset. U.S. Geological Survey Eros Data Center. URL: http://edcdaac.usgs.gov/gtopo30/gtopo30.html.

Graumlich, L.J. 1987. Precipitation variation in the Pacific Northwest (1675 – 1975) as reconstructed from tree rings. Annals of the Association of American Geographers 77:19–29.

Graumlich, L. J., and L. B. Brubaker. 1986. Reconstruction of annual temperature (1590 – 1979) for Longmire, Washington, derived from tree rings. Quaternary Research 25:223–234.

Henderson, J. 2001. The PNV model - an environmental gradient model for predicting elements of the environment and potential natural vegetation across a landscape. U.S.D.A. Forest Service Draft Report on file at Mt. Baker Snoqualmie National Forest.

Impara, P. C. 1997. Spatial and temporal patterns of fire in the forests of the central Oregon Coast Range. Ph.D. dissertation, Oregon State University, Corvallis, Oregon.

Interactive Data Language version 5.2. 2000. Computer software. Research Systems, Inc. Windows NT.

Kertis, J., and M. Huff. 1997. Unpublished data on fire history in the crest of the Cascade Range.

Kertis, J., E. Heyerdahl, and K. Pendergrass. 1995. Unpublished data on fire history in the Augusta Creek watershed.

Kiilsgaard, C. 1999. Land cover type descriptions: Oregon gap analysis 1998 land cover for Oregon. Documentation of GAP2 vegetation dataset.

Long, C. J., C. Whitlock, P.J. Bartlein, and S.H. Millspaugh. 1998. A 9000-year fire history from the Oregon Coast Range, based on a high-resolution charcoal study. Canadian Journal of Forestry Research 28:774–787.

Mock, C. J. 1996. Climatic controls and spatial variations of precipitation in the western United States. Journal of Climate 9:1111–1125.

Morrison, P.H., and F.J. Swanson. 1990. Fire history and pattern in a Cascade Range landscape. U.S.D.A. Forest Service General Technical Report PNW-254.

National Interagency Fire Center. 2000. Total number of wildland fires and acreage burned from January 1 to October 16, 2000 by state. URL: http://www.nifc.gov.

Northwest Interagency Coordination Center. 2000. Unpublished data on annual acreage burned on lands of the U.S. Forest Service, Bureau of Land Management, Oregon Department of Forestry, Bureau of Indian Affairs, Fish and Wildlife Service, and National Park Service, 1990 – 1999.

Oregon Department of Forestry. 2000. Unpublished data on annual acreage burned on lands of the Oregon Department of Forestry, 1985 – 2000.

Oregon GAP Analysis 2 Metadata. 1998. Dataset documentation. URL: ftp://ftp.sscgis.state.or.us/pub/data/statewide/k100/orgapvegt.txt.

Prentice, I. C., W. Cramer, S.P. Harrison, R. Leemans, R.A. Monserud, and A.M. Solomon. 1992. A global biome model based on plant physiology and dominance, soil properties and climate. Journal of Biogeography 19:117–134.

Pyne, S. J. 1982. Fire in America: a cultural history of wildland and rural fire. Princeton University Press, New Jersey.

S-Plus 2000 Professional Edition Release 2. 2000. Computer software. Mathsoft. Windows NT.

Swetnam, T. W., C.D. Allen, and J.L. Betancourt. 1999. Applied historical ecology: Using the past to manage for the future. Ecological Applications 9:1189-1206.

Sykes, M. T., I.C. Prentice, and W. Cramer. 1996. A bioclimatic model for the potential distributions of north European tree species under present and future climates. Journal of Biogeography 23:203–233.

Thompson, R. S., K.H. Anderson, and P.J. Bartlein. 1999. Atlas of relations between climatic parameters and distributions of important trees and shrubs in North America. U.S. Geological Survey Professional Paper 1650 A&B.

Weisberg, P.J., and F.J. Swanson. In press. Regional synchroneity in fire regimes of western Oregon and Washington, U.S.A. Forest Ecology and Management.

Weisberg, P., and F. J. Swanson. 2001. Fire dating from tree rings in western Cascades Douglas-fir forests: an error analysis. Northwest Science 75(2):145-156.

Weisberg, P. 1998. Fire history, fire regimes, and development of forest structure in the central western Oregon Cascades. Ph.D. dissertation, Department of Forest Science, Oregon State University, Corvallis, Oregon.

Weisberg, P. 1997. Fire history of the Bear-Marten watershed. Final report to the U.S.D.I. Bureau of Land Management, Eugene District.

Weisberg, P. 1995. Unpublished manuscript. Fire history in the foothills of the central western Oregon Cascades, U.S.A.

U.S. Geological Survey Gap Analysis Program. 2000. Mission statement. URL: http://www.gap.uidaho.edu/About/MissionStatement.htm.

Van Norman, K. J. 1998. Historical fire regime in the Little River watershed, southwestern Oregon. M.S. thesis, Department of Forest Science, Oregon State University, Corvallis, Oregon.

Venebles, W.N. and B.D. Ripley. 1999. Modern Applied Statistics with S-PLUS. Springer, New York.

Wimberley, M.D., T.A. Spies, C.J. Long, and C. Whitlock. 2000. Simulating historical variability in the amount of old forests in the Oregon Coast Range. Conservation Biology 14(1):167-180.