Contradictions and continuities: a historical context to Euro-American settlement era fires of the Lake States, USA

Abstract

Background

The Lake States experienced unprecedented land use changes during Euro-American settlement including large, destructive fires. Forest changes were radical in this region and largely attributed to anomalous settlement era fires in slash (cumulation of tops and branches) following cutover logging. In this study, I place settlement era fires in a historical context by examining fire scar data in comparison to historical accounts and investigate fire-vegetation-climate relationships within a 400-year context.

Results

Settlement era fires (1851–1947) were less frequent than pre-settlement fires (1548–1850) with little evidence that slash impacted fire frequency or occurrence at site or ecoregion scales. Only one out of 25 sites had more frequent settlement era fires, and that site was a pine forest that had never been harvested. Settlement era fires were similar across disparate ecoregions and forest types including areas with very different land use history. Settlement fires tended to burn during significantly dry periods, the same conditions driving large fires for the past 400 years. The burned area in the October 8, 1871, Peshtigo Fire was comprised of mesic forests where fuels were always abundant and high-severity fires would be expected under the drought conditions in 1871. Furthermore, slash would not have been a major contributor to fire behavior or effects in the Peshtigo Fire when logging was still limited to relatively accessible pine forests.

Conclusions

Historical written accounts of fires and settlement era survey records provide a reference point for landscape changes but lack temporal depth to understand forest dynamics. Tree-ring analyses provide a longer (ca. 400 year) context and more mechanistic understanding of landscape dynamics. While settlement land use changes of Lake States forests were pervasive, fires were not the ultimate degrading factor, but rather likely one of the few natural processes still at work.

Resumen

Antecedentes

Los estados de la Región de los Lagos en EEUU experimentaron cambios sin precedentes en el uso de la tierra, incluyendo incendios grandes y destructivos. Los cambios radicales en los bosques en esta región fueron atribuidos mayoritariamente a incendios anómalos durante la colonización, mediante la quema de pilas (acumulación de ramas y parte superior del dosel de árboles), luego de cortas de aprovechamiento forestal. En este estudio, enmarco los incendios en la era de la colonización en un contexto histórico mediante el examen de datos de cicatrices de fuego en comparación con los datos históricos, e investigo la relación fuego-vegetación-clima dentro de un contexto de 400 años.

Resultados

Los incendios de la era de la colonización (1851–1947) fueron menos frecuentes que los fuegos previos a la colonización (1548–1850), con poca evidencia de que las quemas en pilas impactaran en la frecuencia u ocurrencia a nivel de sitio o a escala regional. Solo uno de los 25 sitios examinados tuvo mayores frecuencias de incendios en la era de la colonización, siendo ese sitio un bosque de pinos que nunca había sido talado/cosechado. Los incendios de la era de la colonización fueron similares a lo largo de diferentes eco-regiones y tipos de bosques, incluyendo áreas con muy diferente historia de uso. Los incendios durante la colonización tendieron a desarrollarse durante períodos significativamente secos, las mismas condiciones que llevaron a grandes incendios en los últimos 400 años. El área quemada el 8 de octubre de 1871 en el incendio de Peshtigo, comprendió a bosques mésicos donde el combustible vegetal fue siempre abundante e incendios de alta severidad eran predecibles debido a las condiciones de sequía del año 1871. Además, las quemas en pilas no habrían contribuido mayormente al comportamiento del fuego o sus efectos, dado que la tala estaba todavía limitada a bosques de pino más accesibles.

Conclusiones

los escritos históricos y los relevamientos de registros de fuegos en la era de la colonización, proveen de un punto de referencia para detectar los cambios en el paisaje, pero carecen de profundidad temporal como para entender la dinámica de los bosques. Los análisis de anillos de crecimiento proveen de contextos más amplios (ca 400 años) y un entendimiento más mecanístico sobre la dinámica del paisaje. Mientras que los cambios en el uso de la tierra durante la colonización en los Estados de los Lagos fueron persistentes, los incendios no fueron el factor último de degradación, sino muy probablemente unos de los pocos procesos naturales que todavía perduran.

Introduction

As climate and fire regimes change, new understanding is needed of the inherent resilience of ecosystems and of the implications for human communities and ecosystem services (Hessburg et al. 2019). Following a period of rapid and intense land-use during Euro-American settlement (settlement) and associated ecosystem changes, forests of the Lake States are now more homogeneous, with less species diversity (Schulte et al. 2007) and less spatial and structural complexity (Meunier et al. 2019a) and are less coupled to local climate controls (Thompson et al. 2013), all of which lead to a loss of resilience in forest ecosystems (Drever et al. 2006). It remains unclear what the relative contributions of logging, slash fires, or subsequent fire suppression were to changes in Lake States forests (White and Mladenoff 1994; Rhemtulla et al. 2009). It was assumed that settlement era fires were anomalous, with more frequent and severe fires resulting from cutover slash (White and Host 2008). Early laws were passed to require disposal of slash (Wright Jr. and Heinselman, 2014 ) and the prevention of wildfires at all costs formed the basis of forest conservation in the Lake States. In the modern era, human-induced disequilibrium (e.g., fire exclusion via landscape fragmentation, grazing, and fire suppression) has weakened the relationships among fire, vegetation, and climate (Marlon et al. 2012; Higuera et al. 2015). This limits our understanding of these interacting factors, thus necessitating longer-term historical perspectives (Parks et al. 2017). Despite these limitations, much of our understanding of fire history comes from recent decades and periods when both climate and human activities have undergone rapid and unique transformations (Marlon et al. 2012; Meunier and Shea 2020).

Perhaps nowhere is the rapid and unique transformation of land illustrated better than the > 20 million hectare “Great Cutover” of the Lake States forests. While some logging preceded legal title, this era began primarily after the original General Land Office public land surveys (Stewart 1935) were carried out (Michigan: 1816–1856, Wisconsin: 1833–1866, Minnesota: 1848–1907) and was largely concluded by 1920 after the Lake States were denuded (Stearns 1997; Kates 2001). While the Cutover is sometimes reduced to a relatively discrete, singular event, it occurred in multiple phases over a period of 60 years. The initial phase of cutting, “river lumbering,” was confined to pine forests near rivers and streams as it was unprofitable to haul logs any distance to water (Whitney 1987). It was not until the 1880s that railroad expansion in northern forests opened new areas for logging including more isolated softwood stands and especially hardwoods like maple (Acer spp.), which could not be floated down rivers (Roth 1898; Larson and Larson 2016). After 1880, both the scale and intensity of logging expanded dramatically, ushering in one of the most intense periods of logging the temperate world has ever seen (Marquis 1975; Whitney 1990). The peak harvest of pine was ca. 1890 (in Michigan and Wisconsin, 1900 in Minnesota) and by 1920 harvest of hardwoods exceeded pine (Stearns 1997). Weaver and Clements (1929) stated that the pine-hemlock forest of the Lakes States had been reduced to such small fragments that there were doubts about its former existence (Whitney 1987).

While cutover logging was globally unprecedented in its speed and intensity (Whitney 1987; Williams 1989; Schulte et al. 2007), it was the repeated and often intense slash fires (Zon and Cunningham 1931; Schulte et al. 2007) that followed which “all the authorities agreed” were the primary source of land degradation rather than logging itself (Roth 1905; Curtis 1959; Whitney 1987). In fact, fires at the time were used to justify logging by timber barons who were depicted not as wasteful capitalists, but rather saving forests from waste by fires (Stimson 1910). The sentiment of “more good pine timber was burned than ever reached the sawmills” was handed down as common knowledge among lumbermen at the time (Zon and Cunningham 1931; Dickmann and Leefers, 2003) and subsequently even by researchers examining the profligate logging practices of the Cutover (Fries 1951; Whitney 1987). Fires fed by pine slash were often viewed as the ultimate degradation of Lake States forests leading to long-term structural and compositional changes and loss (Bormann and Likens 1979). Fires were of course associated with land use changes including exorbitantly wasteful logging, but were also often followed with clearing stumps for farmers (19 million pounds of explosives were used by the University of Wisconsin College of Agriculture alone; Gough 1997), nearly universally failed attempts at farming (Whitney 1987), and subsequent fire suppression.

Certainly, large and destructive fires in this period were common, including the deadliest wildfire in American history—the October 8, 1871, Peshtigo Fire in Wisconsin (Gess and Lutz 2003). Named fire events are mostly useful to managers responding to particular conflagrations whereas ecologists tend to describe large fire years. The fires of October 8, 1871, for example, famously destroyed 9 km² of the city of Chicago and killed several hundred people. At the same time, they burned > 10,000 km² and multiple cities (e.g., Holland, Lansing, Port Huron, Manistee) in lower Michigan (Dickmann and Leefers 2003). These fires are sometimes collectively called “the great Midwest fires of 1871.” In Minnesota, the “Great Fires” denote the especially destructive fire years of 1894, 1908, 1910, and 1918 (Haines and Sando 1969; Johnson 2020). During these destructive fire years, tens of thousands of hectares also burned elsewhere in the Lake States. In Wisconsin, for example, 1894 is the year of the “Phillips Fire” which burned over eight counties in late July, destroying the town of Phillips and killing several hundred people. What these settlement fires have in common is that they were, and continue to be, considered anomalies. Interestingly, 150 years after the fact, we still not only attribute the 1871 fires to extraordinary and anomalous conditions but also take solace in the belief that those conditions will not repeat themselves (Frelich 2002; National Oceanic and Atmospheric Administration, National Weather Service, 2021 ). Yet, in the Lake States region, we have only a nascent understanding of fire regimes in relation to vegetation and climate (Meunier et al. 2019a). Thus far, there has been little effort to view settlement era fires in a greater historical and ecological context.

The first systematic attempt to evaluate the problem of slash as a fire hazard was led by the architect of research in the U.S. Forest Service, Raphael Zon. Zon and Cunningham (1931) found that settlement era fires were a result of weather conditions, not slash, and that any impact on fire intensity/severity would have been both localized, since slash covered a small proportion of cutover areas, and short-lived (Cheyney 1939). Yet, slash remained a primary culprit for uncontrolled wildfires, which posed a “serious problem to those concerned with fire control” (Williams 1955), perpetuating the notion that settlement era fires were unprecedented in intensity and frequency (Dickmann and Cleland, 2002, White and Host 2008) and driven primarily by slash (Wells 1968; Brose et al. 2001; Frelich 2002). It was commonly held, for example, that fire seldom swept into “virgin stands” for more than 10 rods in the absence of logging slash (Lorimer 1980).

Although fuel accumulation was blamed for one of the largest and most destructive fires in Michigan in 1881, the same ground had burned just one decade prior in 1871. At the time, the severity of the 1881 fires was blamed in part by dead timber from the 1871 fire and slash building up for decades (Rummel 2003). However, slash buildup over long periods is in direct contradiction to findings by Zon and Cunningham (1931). In semi-arid regions, the importance of fires in constraining fire growth and severity of subsequent fires is unequivocal (Stephens et al. 2012). It follows that repeated fires in dry pine forests in Michigan have been shown to reduce woody debris and the duff layer, generally resulting in less intense subsequent fires (Drobyshev et al. 2008). Yet, the relative importance of fuels (reduction via burning or addition via slash) to fire behavior and effects in more mesic temperate forests, which predominated in the region of the 1881 fires, are less clear (Comer et al. 1995; Steel et al. 2015).

As the environmental historian William Cronon (1983) stated, “an ecological history begins by assuming a dynamic and changing relationship between environment and culture, one as apt to produce contradictions as continuities.” In this study, I attempted to ascertain the contradictions and continuities of our recent fire history for the Lake States, a period of profound changes and importance that we are only beginning to critically evaluate after more than a century. My objective was to evaluate fires of the settlement era in a context of a longer fire chronology and in relation to vegetation and climate using fine scale dendrochronology reconstructions of fire history. Specifically, I evaluated fire frequency and the similarity of fire events between pre-settlement (prior to 1851), and settlement (1851– ca. 1930) periods. I further assessed fire vegetation characteristics for the 1871 Peshtigo Fire in Wisconsin, USA.

Methods

Study area

This study was conducted across Wisconsin (WI) and the Upper Peninsula (UP) of Michigan, USA, spanning a 4-degree latitudinal (43° to 47°N) and 8-degree longitudinal (− 84° to − 92°W) gradient and included eight ecoregions (Fig. 1; Omernik 1987; Omernick and Griffith 2014). Ecoregions are large ecological landscapes with specific biophysical characteristics defined in part by successive glaciation events that have created diverse forest types across the Lake States (Wisconsin Department of Natural Resources, 2015 ; Meunier et al. 2019b). Within these ecoregions, fire history data from remnant and living fire scarred trees were collected in 25 study sites, each with a component of pine (Pinus spp.), but which were otherwise disparate, allowing comparisons of disturbance history among broad geographical regions with varying physiography and forest types. Ecoregions (east to west) were as follows (Wisconsin Department of Natural Resources, 2015 ; Meunier et al. 2019b):

• (1) Northwest Sands (NWS), WI (4628 km²)—a deep, well-drained sand outwash complex with relatively flat to rolling topography, with lakes limited to the southern portion. This ecoregion had a high proportion of pine (> 93% of settlement survey “witness trees”; Meunier and Shea 2020) and pine barrens (northern savannas on infertile soils). Several notable settlement era fires are documented in historical records in this landscape, including a fire in 1854 (Amery, WI to Iron River, WI), the 1894 Phillips Fire (eight counties surrounding Phillips including NWS, WI; Brown 2009), and an 1898 fire (Barron and Polk counties, Wisconsin Historical Society, 2021 ).

• (2) Central Sands (CS), WI (7068 km²)—a glacial lakebed of well-drained outwash sands intermixed with wetlands, but with few natural lakes or rivers and very little topography. Currently, this landscape consists of a high proportion of dry pine-oak barrens and forest. Historically, the region had higher proportions of pine (62% of original land survey witness trees; Mladenoff 2009, Meunier and Shea, 2020), including red and white pines (P. resinosa, P. strobus) in the northwest (Bruce Mound, Wildcat Ridge, Stoney Bluff sites) and jack pine (P. banksiana) in the central portions (Quincy Bluff site) of the ecoregion. A particularly destructive settlement era fire in this landscape was “The Great Marshfield Fire” in June 1887 (Wood and Marathon Counties, WI).

• (3) Blufflands and Coulees (B&C), WI (23,004 km²)—an unglaciated region of silt and loam soils with few natural lakes and wetlands, but complex ridge and valley topography. Within the Bluff and Coulee region, I collected fire scar samples within a red pine relict (Pine Bluff) occurring on steep bluffs with an exposed sandstone outcrop in an otherwise hardwood-dominated landscape that was primarily white oak (Quercus alba) woodland in pre-settlement times.

• (4) Northern Highlands (NH), WI (6821 km²)—a region with rolling topography with pitted-gravelly sands and abundant natural lakes, rivers, and streams. While sandy soils predominate, the soils tend to be more productive in this ecoregion and support dry-mesic mixed conifer forests which comprised ca. 56% of witness trees in the original survey notes. Historically, this landscape was Wisconsin’s greatest pinery, but also contained abundant eastern hemlock (Tsuga canadensis)-hardwoods, as well as aspen-birch (Populous-Betula). Notable settlement era fires include a large fire in 1904, again in 1907–1908, and 1910 (WI Historical Society Wisconsin Historical Society, 2021).

• (5) Northeast Sands (NES), WI (2865 km²)—deep, well-drained outwash sands with level to rolling topography and scattered, small lakes. This landscape had a high proportion of pine forests (62% of settlement survey witness trees). Notable settlement era fires included one in 1900 (Chequamegon Bay and Menominee River). Two sampling sites in this ecoregion (Wolf Lane and Camp Bird) were old growth pine stands that were never harvested.

• (6) The Keweenaw-Baraga Moraines (KBM), UP (2409 km²)—primarily northern hardwood forest including sugar and red maple (Acer saccharum and A. rubrum), yellow birch (Betula alleghaniensis), aspen (Populus), red and white pine, and eastern hemlock. This ecoregion, and sites within it (Bears Den, Pine Bluff Trail), is characterized by broad ridges and steep morainal slopes with well-drained sandy soils. Pine is a scattered but minor component of forests within this ecoregion. A notable fire in this ecoregion was in 1896 which burned > 900 km² leveling the town of Ontonagon to the west (Dickmann and Leefers 2003).

• (7) The Seney-Tahquamenon Sand Plain (SSP), UP (3765 km²)—primarily mixed conifer swamp, muskeg/bog, and patterned peatlands dominated by tamarack (Larix laricina) and black spruce (Picea mariana) on poorly-drained sand lake plains with mixed pine forests (red, jack, and white pines) on sandy ridges and dunes. Ramsey Lake typified this landscape and was a 900-ha patterned fen peatland.

• (8) The Grand Marais Lakeshore (GML), UP (5284 km²)—red and jack pine ridges and dunes along Lake Superior with interior uplands containing scattered northern hardwoods and eastern hemlock-white pine forests. Peatlands occur in poorly drained lacustrine deposits and in glacial kettle depressions. Sites in this ecoregion ranged from > 1200 ha peatland (Betchler Lake), pockets of pine along Lake Superior (Bay View), to more extensive interior forest (Corner Lake, Hwy 13).

Fig. 1

Study sites (n = 25) among eight ecoregions in Wisconsin and the Upper Peninsula of Michigan in the Lake States, USA

Data collection and analysis

Within natural origin red pine-dominated stands, I collected sections from remnant stumps and nondestructively collected partial sections from fire-scarred dead and living trees at 10 cm height (Fritts and Swetnam 1989). Stands were either old-growth or had been harvested in the cutover period (ca. 1860–1920) but had minimal recent disturbance (e.g., post cutover logging) and contained pre-settlement era stumps. Multiple stands comprised sites which ranged in size from small pine stands within hardwood forests to more extensive pine forests for which I made an effort to sample comparable areas (0.12–2.41 ha, μ = 0.76 ha). Two sites were comprised by unharvested old growth whereas most sites were harvested at the time of cutover but subsequently left relatively undisturbed. Samples were collected from both randomly placed plots used in separate analyses of age structure (e.g., Meunier et al., 2019a, b ) as well as opportunistically (Meunier et al. 2019b; Sutheimer et al. 2021). Whether or not fires recorded at study locations were the same as particular fires recorded in historical written accounts cannot be precisely known, but I sought to investigate their correspondence as well as their relative importance described further below.

In the laboratory, I sanded samples until the cellular structure of xylem was clearly visible with magnification (Grissino-Mayer and Swetnam 2000). I used a dissecting microscope to crossdate tree-rings and assigned exact calendar dates to all fire-scars using standard dendrochronology techniques (Speer 2010). I compiled and analyzed fire-scar chronologies for each site using Fire History Analysis and Exploration System software version 2.0 (FHAES, Brewer et al. 2019). Most trees, once large enough to be protected by thick bark, are not injured by low-intensity fires; however, once injured, trees are more susceptible to repeated injury in subsequent fires and considered “recording trees” (Swetnam and Baisan 1996). It follows that fire scars are evidence of trees that were scarred but survived low- to moderate-severity fires. I analyzed mean fire return intervals (MFRI- the average number of years between fires) within sites in which fires occurred on ≥ 10% of recorder trees by settlement and pre-settlement periods framed broadly by settlement patterns using 1850 as the cutoff (pre-settlement ≤ 1850). For post-settlement chronologies, I included the years up to the time that a major disruption of fire events was apparent excluding unusually long MFRI’s associated with fire exclusion. Filtering, based on scarring percentage, provides a relative index of fire size (Farris et al. 2010). A 10% filter, for example, will only consider fire years recorded on ≥ 10% of recorder trees, thus eliminating fires that may have burned only one or few trees in a single lightning strike or localized incident (Meunier et al. 2014). I also calculated Weibull median probability interval (WMPI) which is a less-biased estimator of central tendency with skewed data (Grissino-Mayer 1999), but MFRI are more commonly reported (e.g., LANDFIRE 2016).

To test for differences in historical fire frequency between settlement and pre-settlement periods, I compared MFRI for fires recorded on ≥ 10% of recorder trees between time periods (SigmaPlot Version 12.0, 2010 ). I used two-tailed Student’s t-test to compare MRFI when data were normally distributed with equal variance and nonparametric Mann-Whitney rank sum tests to compare periods when data were not normally distributed. I compared fire frequencies by site (n = 25) and for sites pooled by ecoregion (n = 8) and used MFRI distribution to compare median values and data spread for pre-settlement and settlement periods.

Composite fire intervals (e.g., fires on ≥ 10% recorder trees) are influenced by the number of fire-scarred trees sampled; the more trees sampled the lower the likelihood of missing a fire (Lafon et al. 2017). MFRIs are also influenced by the level of filtering; higher levels of filtering usually result in fewer, larger fire years and longer MFRI. Therefore, I also evaluated temporal variations in fire with a decadal fire index (DFI) which does not depend on compositing and accounts for sample depth (Hoss et al. 2008). DFI is calculated by first summing the number of fires recorded during each decade and dividing by the number of recording trees that are represented during the decade (Lafon et al. 2017). Dividing by the number of recording trees standardizes fire sums and permits comparison of decades with different sample sizes (Lafon et al. 2017). I regressed DFI against time for the entire pre-exclusion period (both settlement and pre-settlement eras) to test for trends (i.e., positive or negative slope) over time. The Cutover generally followed an east-to-west progression (Maine to Michigan and Wisconsin, then Minnesota) with geographic factors playing an important role in the settlement era lumber industry (Stearns 1997). In addition to site trends, I evaluated regionally composited fire occurrence changes to try and detect these broader spatial scale fire patterns.

There is no robust way to compare historical written accounts of settlement era fires to tree-ring based fire scar records. However, to try to ascertain general agreement between historical accounts and fire-scar data, I calculated the proportion of regions that recorded well-known large fire events. This allowed me to evaluate whether particularly large settlement era fires, as depicted from historical written accounts, were also more widespread (higher proportion of ecoregions recording the same fire year) based on fire scar data (Meunier and Shea 2020). I also composited fire history among all sites (659 fire-scar samples) into a single fire chronology and used a more restrictive filter—fires recorded on ≥ 15% of all recorder trees, with a minimum sample size of 10 recording trees—to compare how the largest fire years, reconstructed with fire-scar samples, compared to known settlement era fires. Setting a minimum sample size of 10 recorder trees helps to reduce the influence of early fire years that scar a high proportion of trees due to small sample sizes. I compared the largest fire years (recorded on ≥ 15% of recorder trees for all samples combined) across both pre-settlement and settlement periods with this composited fire history. I used superposed epoch analysis (SEA) to understand climate drivers (average Palmer Drought Severity Index, PDSI, of four grid points across the region; Cook et al. 2007) for both widespread fire years with ≥ 15% rate of scarring across all sites/samples and for notable settlement era fires via historical accounts.

Generally, fire perimeters for settlement era fires are anecdotal and/or lack detailed descriptions of spatial extent and patterns making fire-vegetation characteristics difficult to ascertain. The 1871 Peshtigo Fire is an exception having reasonably well-documented fire perimeter maps (Fig. 2). To evaluate fire-vegetation characteristics in relation to land use changes (e.g., slash), I overlaid the approximate boundaries for the 4856 km² Wisconsin portion of the October 1871 Peshtigo fire and extracted all settlement era General Land Office Public Land Survey (GLO) witness tree records that intersected the Peshtigo fire boundary (Sickley et al. 2001). The survey, which was designed to form the basis of property boundaries and land records during the settlement era, included notes on the species, diameter, and location relative to section corners of two to four witness trees (Stewart 1935). I overlayed the Peshtigo fire footprint on a raster map of GLO witness trees and extracted all witness trees within the boundary of the Peshtigo Fire to evaluate forest composition of that landscape at the time of settlement using ArcMap 10.3 (Environmental Systems Research Institute (ESRI), 2014 ).

Fig. 2

Maps depicting burned areas in the a October 8, 1871 Peshtigo Fire, Wisconsin where fire loss and death toll was greatest (by courtesy of Encyclopedia Britannica, Inc., copyright 2020; used with permission) and b northeastern Wisconsin region fire maps in 1871 (originally published The Sheboygan Press, October 8, 1929).

Results

This study included 659 crossdated fire-scarred tree samples collected among the 25 sites spanning the period from 1548 (the first fire included in analyses) to 1947 when effective fire exclusion was apparent among all sites. The temporal extent of tree-ring records varied among sites but, collectively, I identified 3229 fires from 1548 to 1947. Despite very different landscape contexts and forest composition among ecoregions and sites, prior to the mid-1900s, fires were historically frequent in all locations, with MFRI ranging from 3 to 15 years (μ = 8) for fires recorded on ≥ 10% of recorder trees (Table 1). When considering fires that were recorded on ≥ 25% of recorder trees (larger fire years), fire frequency decreased slightly but were still frequent (MFRI = 5–34 years, μ = 12). In general, fires were most frequent in the Central Sands, Blufflands and Coulees, and Northwest Sands ecological landscapes (MFRI = 4) and longest in the Keweenaw-Baraga Moraines (MFRI = 12) and Seney-Tahquamenon Sand Plain (MFRI = 11).

Table 1 Fire interval metrics (years) for fires recorded on ≥ 10% of recording trees at each site (n = 25) until the time of disruption of fire events were apparent

Among the 25 different sites, fire frequency was significantly different between pre-settlement and settlement periods at only three sites (Sand Lake, Pine Bluff, and Camp Bird; Table 2). Two sites (Sand Lake and Pine Bluff) had significantly shorter MFRI (P < 0.001, P = 0.008 respectively) in the pre-settlement period and one site (Camp Bird) had significantly shorter return intervals (P = 0.046) in the settlement (MFRI = 4) versus pre-settlement period (MFRI = 8). Results were similar by ecoregion. The Northeast Sands was the only region which had more frequent fire in the settlement period (MFRI pre-settlement = 8, settlement = 4; P < 0.001). The Blufflands and Coulees ecoregion (Pine Bluff) had more frequent fire in the pre-settlement (MFRI = 4) than in settlement (MFRI = 9). There were no statistical differences in fire frequency between periods in the other six ecoregions, though intervals were longer during settlement (Fig. 3).

Table 2 Fire history summaries and tests of difference between Pre-Euro-American (pre-settlement, ≤ 1850) and Euro-American settlement (settlement) eras

Fig. 3

Boxplot of median, interquartile ranges, 90th percentiles, and outliers for pre-settlement (≤ 1850) and settlement (> 1850) era fire return intervals by region (n = 8, arranged west to east, Northwest Sands—NWS, Central Sands—CS, Blufflands and Coulees—B&C, Northern Highlands—NH, Northeast Sands—NES, Keweenaw-Baraga Moraines—KBM, Grand Marais Lakeshore—GML, Seney-Tahquamenon Sand Plain—SSP)

When evaluating trends in fire frequency with decadal fire index (DFI), I found no sites that illustrated a significantly positive slope over time. In contrast, six sites had significantly negative slopes, illustrating reduced fire frequency in the settlement period (Table 3). Most sites had no significant changes over time. Similar trends were evident when evaluating DFI by ecoregion. Only the Northwest Sands ecoregion had a positive slope (Fig. 4) but lacked significant differences in DFI over time. Two ecoregions had significantly negative slopes, Bluff and Coulees (P < 0.001) and Central Sands (P < 0.001), but the Bluff and Coulees ecoregion only had one site (Pine Bluff) and therefore would not be expected to change from site level analyses. In general, there was greater evidence for reduced fire frequency in the settlement period, as compared to pre-settlement, at both site (Table 3) and ecoregion (Fig. 4) scales.

Table 3 Summary of regression models analyzing temporal trends in fire activity for 25 different sites using a decadal fire index (DFI, mean number of fire scars per recording tree, per decade)

Fig. 4

Decadal fire indices for sites (n = 25) composited by region (n = 8). Slopes are linear regression lines for the period of first recorded fires to 1930. Only the Bluff & Coulee (B&C) and Central Sands (CS) regions had slopes significantly different than would be expected by chance (P < 0.001)

During settlement, an estimated 2000 km² (half million acres) per year burned (Mitchell and LeMay 1952), making it difficult to define significant settlement fire years. I did not compile an extensive bibliography of written historical accounts of settlement era fires; however, my interest was primarily to document general agreement between well-known historical events and fire scar reconstructed fires. Three of the four largest settlement era fire years reconstructed via fire scars (based on proportion of ecoregions recording a fire year and filtering of composited fire chronology for all samples) coincided with large fires described in historical accounts. These include (from most widespread to least): 1891, 1910, and 1920 (Table 4, Fig. 5). While fires in 1894 were found in half of ecoregions (Table 4), fire scar records indicate that 1895 was a more widespread fire year (Fig. 5). Both 1894 and 1895 were drought years (averaged PDSI = − 2.008, − 2.380 respectively). All but one historically documented fire year, 1931, which was a severe regional drought year (PDSI = − 3.424), were detected by fire-scars. Fire scar evidence in 1871, one of the more destructive settlement era fire years, did not indicate it as one of the more extensive fire years. Notably, fire scar records of this fire year, and all fire years detected by fire-scar methods, are primarily limited to low- and/or moderate-severity fires that scar trees without killing them (Kent 2014). Such records are consistent with more comprehensive maps showing a more extensive fire footprint than the relatively limited area impacted by high-severity Peshtigo Fire (Fig. 2b). Both large fire-scar reconstructed fire years (Fig. 5) and notable fires via historical accounts (Table 4) were significantly related to dry conditions (Fig. 6).

Table 4 Notable Euro-American settlement era wildfires in the Lake States organized by percentage of ecoregions (n = 8) in Wisconsin and the Upper Peninsula of Michigan detected with fire-scars and then chronologically

Fig. 5

Composite fire chronology for all sites of large fire years recorded on ≥ 15% recording trees with ≥ 10 recording trees. Top graph illustrates number of recorder trees (line) and percent of all recorder trees scarred (bars) over time

Fig. 6

Results of superposed epoch analysis (SEA) of the average Palmer Drought Severity Index (PDSI) across the region (Cook et al. 2007) for years prior and subsequent to fire event years (year 0) for a notable historical accounts of settlement era fires (i.e., Table 4) and b ≥ 15% average rate of scarring (minimum 10 recording trees) across all ecological landscapes. Positive PDSI values indicate wet conditions, negative values represent dry conditions. Solid bars indicate PDSI values outside a 95% confidence interval based on 1000 Monte Carlo simulations of random distributions of annual PDSI (a 1850–1950, b 1600–2000)

The Peshtigo Fire is somewhat unique in that there are detailed fire perimeter maps allowing at least qualitative analysis of fire-vegetation-climate relationships. Settlement era witness tree maps of the Peshtigo Fire region indicated that portions of the 1871 fires were a mesic “sugar bush”—a colloquial term for maple (Acer) forest used for syrup production (e.g., Lower, Upper, and Middle Sugarbush; Fig. 2a). The most abundant tree species out of 11,014 witness trees identified in the footprint of the Peshtigo Fire at the time of GLO Public Land Survey (ca. 10 years prior to the fire) was hemlock (Tsuga canadensis; 21%), followed by cedar (Thuja occidentalis; 14%), beech (Fagus grandifolia; 12%), tamarack (Larix laricina; 8%), and sugar maple (Acer saccharum; 8%; Fig. 7). Less than 10% of witness trees in the footprint of the Peshtigo fire were merchantable pine of any kind (most of which was white pine—a common component of mesic hardwoods; Curtis 1959). This region was and remains dominated by mesic forests (Wisconsin Department of Natural Resources, 2015 ) which were not harvested at any scale until a decade or more following the Peshtigo fire (Stearns 1997).

Fig. 7

Percent of all trees (n = 11,014) by species recorded by Euro-settlement era Public Land Survey notes (1832–1866) within the 485,623-ha footprint of the October 8, 1871 Peshtigo fire

Discussion

Understanding disturbance regimes in relation to climate and land use change is vital for fostering resilient ecosystems that can accommodate an uncertain future (Landres et al. 1999). Historical context is needed to interpret past and potential climate change influences on disturbance processes but is reliant on robust historical baseline data. The need for robust baselines is particularly acute for ecosystems with highly altered disturbance regimes (Higgs et al. 2014; Knight et al. 2020). The Lake States have highly altered disturbance regimes (Meunier et al., 2019a, b ) as well as an aberrant settlement reference period when rapid and intense land use changes were pervasive. It was assumed, for example, that settlement era fires were anomalous with more frequent and severe fires resulting from cutover slash (White and Host 2008); yet there is little evidence for slash as a widespread and prolonged fire hazard (Zon and Cunningham 1931) and little or no evidence that fire was more frequent in the settlement period (Tables 2 and 3, Figs. 3, 4, and 5). Fuel type (e.g., forest composition), in addition to fuel loading (e.g., slash), needs to be considered in relation to fire severity, particularly for destructive settlement era fires like the October 1871 Peshtigo fire. The assumption that fires were another (or the) degrading force of forest following exploitative settlement era land-use changes (Johnson 2020) may be fundamentally flawed and formed, in part, by ignorance of basic historical fire regime characteristics in Lake State forests.

In the context of a longer (ca. 400-year) fire history, settlement era fires were not uncharacteristically frequent nor is there an obvious relationship between fire frequency and slash (Zon and Cunningham 1931). Only one out of 25 sites, Camp Bird in the Northeast Sands (NES), had a significantly shorter fire return interval in the settlement period (Table 2, Fig. 3), and that site was a never-harvested old growth pine stand where slash would have been absent. In most cases fires were less, not more, frequent in the settlement era (Tables 1 and 2, Figs. 3 and 4). Interestingly, many of the same settlement era fires attributed to slash are documented in unharvested portions of the Boundary Waters Canoe Area Wilderness (BWCAW; Heinselman 1973) and also the 95,313 ha Menominee Indian Reservation (Sands and Abrams 2011). In spite of never having had exploitative logging on the Menominee Reservation, Sands and Abrams (2011) documented 36 settlement era fires between 1850 and 1910, 14 of which are synchronous with notable historical accounts of settlement fires in the region: 1854, 1864, 1868, 1871, 1879, 1881, 1887, 1891, 1893, 1894, 1896, 1898, 1900, and 1910 (Table 4). Settlement fires in the “virgin” BWCAW forests are also very similar to elsewhere in the Lake States region (e.g., 1854, 1863-1864, 1868, 1871, 1874-1875, 1880-1881, 1887, 1894, 1900, 1910, 1918, 1920, 1936; Heinselman 1973). Importantly, settlement era fires in the BWCAW are generally not viewed as anomalous, in part due to the foundational fire history work by Heinselman (1973), which provided greater context to the incidence of fire there.

Several studies have suggested that the combination of logging and slash fires followed by fire suppression in the 1930s drastically altered forest composition in the Lake States (Miller et al. 2010) and led to a loss of coniferous species and increase in deciduous species in northern Wisconsin (Mladenoff and Howell 1980; White and Mladenoff 1994; Radeloff et al. 1999; Steen-Adams et al. 2007). However, it remains unclear what the relative contributions of logging, slash fires, or subsequent fire suppression were to these changes (White and Mladenoff 1994; Rhemtulla et al. 2009). Cook (2000) found that tree establishment dates in Lake States pine-oak forest peaked following logging and fires, particularly between 1891 and 1911, which include some of the most widespread settlement era fire years (e.g., 1891, 1894-1895, and 1910; Table 4, Fig. 5). Close coupling between tree recruitment and low-severity fires, including in the settlement period, was also noted in the BWCAW (Heinselman 1973). In a study in western Oregon and Washington, effects of slash burning following clearcut harvesting had little effect on the quantity of natural stocking of conifers and subsequent hazard rating was lower on burned than unburned plots (i.e., lower spread rate and more easily controlled; Morris 1970). Remarkably, while land use history varied across the disparate forests and ecoregions in this study (e.g., different composition, harvested or not, timing and intensity of harvest), two factors are common among all locations: fire was historically frequent, and fire was universally excluded by the mid-twentieth century (Table 1). It is likely that 100 years of fire suppression, rather than a period of more frequent fires, were primary drivers of expansive forest changes (Paulson et al. 2016).

It seems that fires in the Lake State were influenced by abundant fuels and, periodically, extended drought conditions. The relationship between settlement land use changes and fire severity remains an unresolved issue; however, it is known, at least in theory, that many Lake State conifer forests inherently have heavy fuel loads (canopy and surface fuels) where fire behavior is controlled more by weather at the time of fire than by any other factor (Bessie and Johnson 1995; Frelich 2002). Cutting (and windthrow) can change fuel moisture of slash, which can effect flame length and potentially increase fire severity (Evans and Wright 2017). Fuel moisture of slash increases with compaction over time which acts counter to combustion dynamics (Wright et al. 2019). Fuels also tend to decompose much faster in eastern forests, with higher productivity (moisture and temperature) than in most of the western US (Graham and McCarthy 2006) where fuel accumulation increases risk of high intensity fire. Notably, the 1871 fires in Wisconsin had been burning for many weeks prior to October with little rain since the early part of July and below average snowfall in winter of 1870-1871 (Gess and Lutz 2003). Residents described a constant crackling of fire at night and smoke was omnipresent (Pernin 1971). These fires were primarily lower severity, much of which burned pine forests to the north of the high severity October 8 Peshtigo Fire (e.g., northern Oconto and Marinette Counties), where presumably slash would have been present (Fig. 2b). These low severity fires of 1871 were detected across study sites in Wisconsin (NWS, CS, NES), the Menominee Indian Reservation to the northeast (Sands and Abrams, 2011), and in the Huron Mountains in the Upper Peninsula of Michigan (Muzika et al. 2015).

Schulte and Mladenoff (2005) in their analysis of GLO data found that stand-replacing fires were never recorded by surveyors within the ecological landscapes associated with the October 8, 1871 Peshtigo Fire (Fig. 2a). Wildfire frequency and intensity are broadly inversely related (Pickett and White 1985; Turner et al. 1989). Fire regimes in many forested systems can be categorized as ranging from “fuel-limited” to “climate-limited,” where the former are characterized by more frequent, lower-severity wildfires and the latter by infrequent, more severe wildfires (Steel et al. 2015). The mesic forests of the Peshtigo Fire region (Fig. 7) would be expected to burn infrequently and with high severity. It is not surprising that GLO survey data would not detect fires in this landscape which would be characterized by infrequent fires nor is it surprising that this landscape would burn with the intensity and devastation that it did with abundant natural fuel loads (e.g., hemlock, cedar, larch forests). In 1871, logging was still restricted to river lumbering of pine, and while land clearing was associated with settlement, it would likely have been limited in comparison to the ca. 5000 km² area burned. Very little logging had occurred within the Peshtigo Fire footprint, and slash from logging would not have been abundant there in 1871. Regardless of the contributions of settlement activities to the Peshtigo Fire, high-severity fires would be expected to predominate in forests there. Notably, the high-severity Peshtigo Fire represents a relatively small proportion of area embedded within a larger burn area that burned as low-severity fire in that same year (Fig. 2b).

In 1871, there was prolonged regional drought (average PDSI 1871 = − 2.364, 1870 = − 2.470) where fires had been burning for more than a month, which were then fanned by strong southeasterly winds (National Oceanic and Atmospheric Administration, National Weather Service, 2021 ) into an abnormally dry forest with abundant fuels. This story is not unique to the Peshtigo Fire or to 1871, but rather it was the way many large fire years burned over the last 400 years in the Lake States (Figs. 5 and 6). While land use changes in relation to wildfires occurring during settlement in the Lake States cannot be discounted, they also cannot explain scores of similar fire years, over widely varying landscapes, and over centuries (Fig. 5).

Historical reference information is valuable for understanding how current conditions arose, as well as for developing a mechanistic understanding of ecosystem dynamics and predicting future conditions (Safford et al. 2012; Miller and Safford 2017). Historical data, like written accounts at the time of Euro-American settlement, or settlement era Public Land Survey data, can be useful in describing the magnitude of change in conditions relative to reference landscapes (Rhemtulla et al. 2009). However, such data are not always useful in assessing the drivers of change or in forecasting future conditions (Frelich 2002). Likewise, a single “settlement” reference is a static view of ecosystems that were and remain inherently dynamic. What becomes apparent when considering a longer history and greater context is that frequent fires were one of few continuities within forests of the Lake States during settlement. This is a starting point for unraveling the effects of the myriad degrative land uses in the settlement period and for understanding the contradictions of defining forest conservation in the Lake States on the premise of fire suppression. Although the nature of future forests will be determined in part by climate change and other exogenous variables, land use and fire have been and are likely to remain the driving factors in shaping forests (Rhemtulla et al. 2009). The question is what type of land uses and what type of fire will be part of those future forests.

Conclusions

Settlement era fires (ca. 1851–1947) were believed to have been abnormally frequent and primarily a result of slash build up following extensive cut-over of Lake States forests. However, historically (over the last 400 years), fires were always relatively frequent, occurring on average every 8 years (MFRI = 3–15 years). Settlement fires in the Lake States were not simply slash fires, but a combination of weather, climate, and fuel arrangements. Large fires tended to correspond to extended drought conditions regardless of time period. Destructive fires in the Lake States, like the October 8, 1871, Peshtigo Fire, burned in mesic forests where large volumes of pine slash would not have been present; however, large volumes of fuels were always present and where fires would be expected to burn with high severity.