A S ummAry of f ire f requency e StimAteS for c AliforniA V egetAtion before e uro- A mericAn S ettlement

California fire regimes have been altered from those that occurred prior to Euro-American settlement, and are predicted to continue to change as global climates warm. Inclusion of fire as a landscape-level process is considered essential to successful ecological restoration in many ecosystems, and presettlement fire regimes provide foundational information for restoration or “realignment” of ecosystems as climate change and land use changes progress. The objective of our study was to provide an up-to-date, comprehensive summary of presettlement fire frequency estimates for California ecosystems dominated by woody plants, and to supply the basis for fire return interval departure (FRID) mapping and analysis in California. Using the LANDFIRE Biophysical Settings (BpS) vegetation-fire regime types as a framework, we used literature review and the outcomes of regional expert workshops to develop twenty-eight presettlement fire regime (PFR) groups based on similarity of their relationships with fire. We then conducted an exhaustive review of the published and unpublished literature pertaining to fire return intervals (FRIs) observed prior to significant Euro-American settlement in the twenty-eight PFRs, and summarized the values to provide a single estimate of the mean, median, mean minimum, and mean maximum FRI for each PFR. and their interactions; and to provide data upon which models of “properly function-ing” or “resilient” ecosystems might be built (Wiens et al. , in press). Drawing comparisons between presettlement and current fire regimes can also assist land managers in prioritizing areas for ecological restoration. Fire return interval departure (FRID) analysis facilitates quantification of the difference between current and presettlement fire return intervals (FRIs), allowing managers to target areas at high risk of type conversion due to altered fire regimes (Caprio et al. 1997, Caprio and Graber 2000, van Wagtendonk et al. 2002). Robust estimates of the variability of presettlement FRIs in different vegetation types are a crucial part of FRID analysis, yet most fire history studies are highly localized, making it difficult to apply the results of individual studies to a regional-scale mapping and assessment effort. should be a useful reference for and whether are seeking a general estimate of the central tendency and variability of in a broadly defined vegetation type, background information for a planned restoration project or a mechanistic model of vegetation-fire interactions, or a list of literature pertaining to a specific vegetation type or geographic location.


introDuction
Fire is an important process in many of California's ecosystems, and it is becoming increasingly evident that fire regimes (including fire frequency, severity, extent, spatial patterning, etc.) have been greatly altered in some vegetation types by land use patterns and altered ecosystem processes associated with Euro-American settlement (i.e., after 1850) (Sugihara et al. 2006. Climatic variability at a variety of temporal scales has been shown to be associated with fire regime fluctuations in the past (Swetnam 1993). Anthropogenic climate change is a driver of current observed trends of increasing fire activity, and is predicted to continue to alter fire regimes and vegetation types in the future (Lenihan et al. 2003, Westerling et al. 2006, Gedalof 2011, National Research Council 2011. Consideration of fire as a landscapelevel process is considered essential to facilitating ecological restoration or "realignment" (sensu Millar et al. 2007) efforts intended to increase ecosystem resilience in the face of climate change (North et al. 2009a). Restoration of narrowly defined historical conditions may no longer be a preferred management prescription in light of the uncertainty surrounding the effects of climate change on fire and other ecological processes. However, information on fire regimes before Euro-American settlement is of fundamental importance to modern and future management of many ecosystems in western North America (Millar et al. 2007;Wiens et al., in press). Such historical information can help, among other things, to document the current status of fire in ecosystems and trends in fire activity and ecological effects over time; to encourage understanding of the underlying mechanisms that drive ecosystem response to changes in climate, fire, landscapes, and their interactions; and to provide data upon which models of "properly functioning" or "resilient" ecosystems might be built (Wiens et al., in press).
Drawing comparisons between presettlement and current fire regimes can also assist land managers in prioritizing areas for ecological restoration. Fire return interval departure (FRID) analysis facilitates quantification of the difference between current and presettlement fire return intervals (FRIs), allowing managers to target areas at high risk of type conversion due to altered fire regimes (Caprio et al. 1997, Caprio andGraber 2000, van Wagtendonk et al. 2002). Robust estimates of the variability of presettlement FRIs in different vegetation types are a crucial part of FRID analysis, yet most fire history studies are highly localized, making it difficult to apply the results of individual studies to a regional-scale mapping and assessment effort.
Much work has been accomplished in documenting historical fire regimes in various should be a useful reference for scientists and resource managers, whether they are seeking a general estimate of the central tendency and variability of FRIs in a broadly defined vegetation type, background information for a planned restoration project or a mechanistic model of vegetation-fire interactions, or a list of literature pertaining to a specific vegetation type or geographic location.
Keywords: California, fire frequency, fire history, fire return interval, FRID, presettlement fire regime Citation: Van de Water, K.M., and H.D. Safford. 2011. A summary of fire frequency estimates for California vegetation before Euro-American settlement. Fire Ecology 7(3): 26-58. doi: 10.4996/fireecology.0703026 vegetation types throughout California, and while several manuscripts summarize different subsets of this information, they are often either restricted to data derived from tree-ring studies, limited to forest vegetation types in a particular geographic region, or not intended to be comprehensive (e.g., Heyerdahl et al. 1995). Thus, scientists and land managers currently lack a single source that summarizes all of the literature pertaining to presettlement fire regimes. The objectives of this paper are to provide an up-to-date comprehensive summary of presettlement fire frequency estimates for California ecosystems dominated by woody plants, and to provide the quantitative basis for fire return interval departure (FRID) mapping across California.

metHoDS
Although the state of California is home to a high diversity of species, vegetation types, and fire regimes (Barbour et al. 2007), similarities among fire regimes and their effects on vegetation generally allow the organization of ecosystems into broad fire regime groups. Published efforts to categorize relationships between fire and vegetation in California include Agee (1993; northern California), Skinner and Chang (1996; Sierra Nevada), Stephenson and Calcarone (1999; southern California), Arno (2000; western US), Sugihara et al. (2006;statewide), Sawyer et al. (2010;statewide), and the LANDFIRE project (2010; Rollins 2009; entire US). Both the Sugihara et al. (2006) and Sawyer et al. (2010) efforts drew from a series of Joint Fire Science Program supported regional workshops held between 2000 and 2002 that reunited fire and vegetation experts from across the state and developed descriptions of fire regime characteristics for California vegetation communities. All of this information also fed the development of the LANDFIRE Biophysical Settings (BpS), which are potential natural vege-tation (PNV) types linked to quantitative models of disturbance and succession (Rollins 2009). The disturbance-succession models for the California BpS types (which apply to the pre-Euro-American settlement period) were developed at a series of regional expert workshops sponsored by The Nature Conservancy in 2004 and 2005. After refinement and peer review, the BpS classification was finalized and mapped as part of the national LANDFIRE project (see Rollins 2009 and www.landfire. gov for details).
As an evolutionary outgrowth of the previous fire regime work cited above, the LAND-FIRE Biophysical Settings represents the current state of the art for linking vegetation and pre-Euro-American settlement fire regimes across California. There are more than eighty individual BpS types mapped in California by LANDFIRE, but some are extremely uncommon, and many share similar fire regimes. Using the fire regime information provided for each BpS in the type description (LANDFIRE 2010), and referring to integrative vegetation and fire resources in the literature (see citations above, plus, e.g., Burns and Honkala 1990, Potter 1998, Barbour and Billings 2000, Barbour et al. 2007) and on the internet (e.g., the Fire Effects Information System [http://www. fs.fed.us/database/feis/]), we reduced the BpS list down to a smaller number of pre-Euro-American settlement fire regime groups (PFRs) that we subjectively considered sufficiently different to warrant retention. From our research, we also identified a number of PFRs that were not represented in the BpS classification. PFRs were designed to balance a reasonably small number of fire regime groups with sufficiently high discrimination in fire regime characteristics. We solicited peer review of the PFR list from 27 California fire and vegetation ecology experts and received responses from eleven. After adjustment, our final list included 28 PFRs.
For each PFR, we conducted an exhaustive review of the published and unpublished liter-ature pertaining to mean, median, minimum, and maximum fire return intervals observed prior to significant Euro-American settlement (i.e., the middle of the nineteenth century). Sources included fire histories derived from dendrochronological and charcoal deposition records, modeling studies, and expert quantitative estimates. Priority was given to studies conducted in California, but sources from other states in western North America were included as appropriate for PFRs for which information was limited. When all sources were compiled, the average was taken of all mean, median, minimum, and maximum FRI values to yield a single mean, median, mean minimum, and mean maximum FRI estimate for each PFR. Thus, the minimum and maximum FRI estimates we provide for each PFR are not absolute minima and maxima, but typical mean values that would be expected across the geographical range of each PFR.
For conifer-dominated PFRs, most FRI values considered in this assessment were derived from small-scale (<4 ha) composite dendrochronological fire histories including records from multiple trees in a defined area, although some values were obtained from modeling or stand age-based studies (in the latter case, for PFRs characterized by stand-replacing fires). Composite FRIs often represent the fire history of a given area better than point FRIs (derived from a single tree) because some fire events fail to scar every recording tree within the fire perimeter, especially in regimes characterized by frequent low intensity fire Stephens 2007, Falk et al. 2011). Furthermore, composite FRIs are more sensitive and better suited to analyzing changes in fire occurrence than point FRIs (Dieterich 1980, Swetnam andBaisan 2003). While there is some variability introduced by using composite FRIs from different-sized areas, they are less likely to underestimate presettlement FRI values than point FRIs. Table 1; characteristic dominant woody species for each PFR are listed in Table 2. Four PFRs were not represented by any BpS types, due to their geographic rarity or their focus on single species. These were the "fire sensitive spruce or fir," the "big cone Douglas-fir," the "shore pine," and the "silver sagebrush" PFRs.

Relationships between the PFRs and LANDFIRE BpS types are shown in
We derived fire frequency estimates for the 28 PFRs from 298 sources ( Table 2). Most of our sources (213 of 298; 71.5 %) were based on data collected in California. For the average PFR, 26.5 % of sources were non-Californian, but seven PFRs had more than 50 % of their sources from outside California (Table 2). These seven, which accounted for about two thirds of all of the non-California sources, are PFRs for which the dominant woody species are at the southern or western edge of their range in California, or are California endemics and very rare in the state (e.g., Abies bracteata [D. Don] D. Don ex Poit and Picea breweriana S. Watson in the fire sensitive spruce or fir PFR). Sixteen PFRs had ≤20 % of their sources from outside California, and seven had exclusively California sources.
Derived mean, median, mean minimum, and mean maximum fire frequencies for each PFR are given in Table 2. Information on median FRIs was lacking for some PFRs, so median values were either taken from expert quantitative estimates of mean FRI (desert mixed shrub, semi-desert chaparral) or were not estimated (coastal fir, shore pine, sprucehemlock). Because FRI distributions are often skewed (with more short or long intervals, depending on the PFR), median FRI values may be a better approximation of how often a given PFR burned than mean FRIs (Falk 2004).

DiScuSSion
Our summary of California's presettlement fire regimes should be a useful reference for scientists and resource managers, whether they are seeking a general estimate of the central tendency and variability of FRIs in a broadly defined vegetation type, background information for a planned restoration project or a mechanistic model of vegetation-fire interactions, or a list of literature pertaining to a specific vegetation type or geographic location. A high degree of confidence can be placed in the validity of the FRI values for most conifer PFRs, especially in the Sierra Nevada, due to the abundance of published dendrochronological studies. Less confidence is afforded to the FRI values of PFRs for which presettlement fire history is less well-studied, such as California juniper, desert mixed shrub, semi-desert chaparral, and silver sagebrush. For shrubdominated PFRs in which presettlement fires are difficult to detect due to a lack of dendrochronological evidence, FRI values were derived from other types of data that may be less precise, such as charcoal in sediment cores, modeling, and expert quantitative evidence.
More research is needed in PFRs that currently have little quantitative fire history data available for California (see Table 2), or have high geographic variability in FRIs. The difficulties associated with obtaining high-resolution presettlement FRI data in shrub-dominated vegetation types categorically necessitates further study in most of these PFRs (e.g., big sagebrush, black and low sagebrush, chaparral, coastal sage scrub, curl-leaf mountain mahogany, desert mixed shrub, montane chaparral, semi-desert chaparral, silver sagebrush), and perhaps innovation of new or adaptation of ex- isting fire history techniques. Similarly, PFRs dominated by tree species that are easily killed by fire (California juniper, coastal fir, fire sensitive spruce or fir, pinyon-juniper, shore pine, spruce-hemlock) require further study and application of techniques other than fire scar studies. The PFRs of limited geographical distribution in California (bigcone Douglas-fir, coastal fir, fire sensitive spruce or fir, Port Orford cedar, shore pine, spruce-hemlock, western white pine) are chronically understudied. Other PFRs (shore pine, desert mixed shrub, spruce-hemlock, California juniper, coastal fir, pinyon-juniper, western white pine, subalpine forest) are characterized by high geographic variability in fire frequency (high standard error of FRI statistics), requiring scientists and managers to carefully search for literature from local or similar areas.
Several interesting patterns in FRIs within and among different PFRs emerged from the body of fire history literature assessed for this article. For example, analyses of the correlation between fire scar sampling area and fire return interval revealed no trend of decreasing FRI with increasing sampling area for all PFRs pooled and most PFRs individually. Sampling area was significantly correlated with mean minimum FRI for the big sagebrush (r = 0.867, P = 0.012), Port Orford cedar (r = −0.974, P = 0.026), and red fir (r = 0.742, P = 0.014) PFRs. The trend of decreasing FRI with increasing sampling area for the Port Orford cedar PFR was consistent with established expectations (Baker andEhle 2001, Swetnam andBaisan 2003), while the opposite trend for the big sagebrush and red fir PFRs may be indicative of the long minimum return interval, mixed severity, and stand replacement fire regimes that typify these vegetation types (Sugihara et al. 2006).
Ignitions by indigenous peoples were likely a large component of the presettlement fire record in some PFRs, such as redwood (Greenlee and Langenheim 1990) and oak woodland, and are difficult or impossible to definitively differentiate from lightning ignitions, although fire cause may be inferred from seasonality in some cases (Anderson and Moratto 1996). Some vegetation types in certain areas were probably maintained mostly by presettlement anthropogenic fire regimes, which may have resulted in vegetation type conversions in some parts of the landscape prior to Euro-American arrival. Widespread indigenous ignitions were probably uncommon in other PFRs, however, such as subalpine forest and desert mixed shrub (Anderson 2005). Regardless, no attempt is made in this assessment to differentiate between lightning and indigenous ignitions.
This paper provides background information for the FRID mapping products developed by the Forest Service's Pacific Southwest Region Ecology Program and Remote Sensing Lab ; available at: http:// www.fs.fed.us/r5/rsl/clearinghouse/r5gis/ frid/). These annually updated maps provide information on geographic distribution of PFRs, and a number of different FRID statistics calculated using the California fire perimeters database (available at: http://www.frap. cdf.ca.gov/data/frapgisdata/select.asp). These map layers are useful for land and resource planning and assessment, as well as other natural resource applications such as fuels treatment planning, postfire restoration project design, management response to fire, and general ecological understanding of the historical and current occurrence of fire on the California national forests and neighboring jurisdictions.
Our process necessarily generalized across scales of both space and time. In general, and assuming all else is equal, areas with higher precipitation or less ignition within a given PFR would be expected to burn less often than drier areas with an ignition source (Agee 1993, Sugihara et al. 2006). A PFR in northwestern California therefore might be expected to support somewhat longer fire return intervals than the same PFR in southern California. A solution for this may be to use the median fire fre-quency as the preferred measure of central tendency for PFRs in parts of their range where vegetation is relatively more flammable, and the mean fire frequency where vegetation is relatively less flammable (at least where the median is shorter than the mean, which is the typical case). Patch sizes can also influence fire frequency, with small patches of mesic vegetation embedded in a matrix of drier vegetation having shorter fire return intervals than large patches of mesic vegetation, and viceversa (Agee et al. 1990a). Obviously, where higher local accuracy is required, the reader should consult the primary literature (e.g., see the citations supporting Table 2).
Temporally, changes in vegetation on California landscapes since the middle of the nineteenth century can make comparisons between historical and contemporary conditions difficult. A good example is provided by the geographic distribution of the yellow pine PFR, which is dominated by ponderosa pine (Pinus ponderosa C. Lawson) and Jeffrey pine (P. jeffreyi Balf.). The Forest Service mapped vegetation on about 60 % of its California lands in the 1930s (Wieslander 1935). Modern vegetation mapping can be generalized to the same polygon resolution and compared with the 1930s maps to get a broad idea of landscapelevel vegetation changes. After >75 years of fire exclusion, logging, and other anthropogenic change, the area occupied by the yellow pine PFR appears to have decreased by about two thirds in the central Sierra Nevada, with about two thirds of the loss due to infilling by shadetolerant (mostly fire-intolerant) conifer species, for example from the genus Abies (Thorne et al. 2008; J. Thorne, University of California, Davis, USA, and H. Safford, USDA Forest Service, Vallejo, California, USA, unpublished data). The FRID mapping is often based on contemporary vegetation data, and these sorts of temporal changes cannot be properly accounted for. After completion of digitization of the 1930s vegetation maps, we hope to use them (where they are available) to update the geographic distribution of PFRs to allow a more accurate assessment of changes in fire frequency.
Although this study presents summarized estimates of presettlement fire frequency, it does not imply that contemporary fire should necessarily be applied at historical intervals, which may be neither feasible nor desirable in the context of altered anthropogenic influences and climatic conditions (Anderson and Moratto 1996;Millar et al. 2007;Wiens et al., in press). Instead, the estimated presettlement FRIs are intended to serve as an assessment tool for comparison with current fire regimes and trends in those regimes, and to provide general guidelines for ecological restoration (or realignment) in vegetation types that are currently in jeopardy of type conversion due to fire frequencies that are outside the historical range of variation. In order to promote ecosystem resilience in the face of climate change and other uncertainties, efforts to restore fire to ecosystems should focus on the variability of fire frequencies (and other characteristics of the fire regime) that historically facilitated resilience, rather than applying fire to an ecosystem precisely at the mean or median interval.

AcKnoWleDgmentS
We thank C. Skinner, C. Chang, S. Stephens, R. Martin, N. Clinton, E. Heyerdahl, D. Berry, J. Agee, T. Keeler-Wolf, J. Sawyer, and all of the authors of species reports in the Fire Effects Information System for their previous work compiling fire history literature, which greatly aided our efforts. We also thank D. Schmidt for developing the first generation FRID maps for southern California and supporting metadata, which formed a basis for this paper. Funding for this project was provided by the Forest Service, Pacific Southwest Region Ecology Program. Brunelle, A., and C. Whitlock. 2003. Postglacial fire, vegetation, and views 24: 2281-2300. doi: 10.1016/j.quascirev.2004.11.010 Buechling, A., and W.L. Baker. 2001