Humans,fire,and ecology in the southern Missouri Ozarks,USA

Study area

The Ozarks are a dissected highland biogeographic region located primarily in southern Missouri and northern Arkansas (35–39°N lat.), bordered by the Missouri River to the north and the Arkansas Valley to the south (Figure 1a; King, 1973). The Ozarks are the westernmost extension of the Eastern Deciduous Forest, with prairie to the west. Today, the Missouri Ozarks are dominated by Quercus-Carya forest (Braun, 1950; Dyer, 2006; Hanberry et al., 2014b; King, 1973). Pinus echinata was abundant in southeastern Missouri, until it was heavily logged at the end of the 19th century (Figure 1c; Braun, 1950; Dyer, 2006; Holmes, 1968).

Archaeological investigations from the Ozarks suggest that human occupation of the region was transient throughout the Paleoindian (~10,950–9550 BP), Archaic (~9550–2500 BP), and Woodland (~2500–1050 BP) periods (Klippel et al., 1978; Ray et al., 1998; Wolverton, 2005; Wood and McMillan, 1976). Although the Mississippi and Ohio River Valleys were heavily populated by the agricultural Mississippian societies after 900 CE, 1050 cal. yr BP (peaking by ~1120 CE); the Ozarks were not heavily occupied until major Euro-American settlement in the 19th century CE (Bird et al., 2017; Holmes, 1968; Pauketat, 2003). Archaeological and historical evidence suggests that the Osage tribe of Native Americans arrived in the region between the end of the 15th and the beginning of the 16th centuries CE (Bailey, 2001; McMillan, 2014b) (Figure 1a). The Osage subsisted on mixed hunter-gatherer-horticulturist strategies, living primarily in permanent villages on the prairie-forest border (Bailey, 1973). Osage hunting parties were short, seasonal ventures into the prairie for bison and into the Ozark woodlands for elk and deer, often utilizing fire in the process (Ponziglione, 1882; Schoolcraft, 1821, 1853).

Sweeton Pond is a small, deep sinkhole pond (36.7914°N, 92.2231°W, 304 m elev., 0.5 ha, 14.6 m maximum depth) in northern Ozark County, Missouri (Figure 1). The pond lies in a woodlot dominated by Quercus alba, Carya, Ulmus, Nyssa sylvatica, and Juglans nigra. Cephalanthus occidentalis grows in the littoral zone and Salix nigra is present along the shore. The pond is located near the extreme northwestern limit of P. echinata, which has a modern range in the warm, humid southeastern United States, where January temperatures range from 0°C to 12°C and July temperatures range from 25°C to 30°C, and annual precipitation ranges from 102 to 152 cm·yr⁻¹ (Lawson, 1990; Webb et al., 1993; Williams et al., 2006). The species is shade intolerant and highly dependent on fire for regeneration (Guyette et al., 2007; Lawson, 1990). Locally, Pinus was concentrated in the valleys of the North Fork River to the east of Sweeton Pond and Bryant Creek to the west until ~1870 CE (Figure 1c; Holmes, 1968).

Public Land Surveys (PLS) vegetation reconstruction

The original PLS from Missouri provide a record of tree distributions from 1815 to 1850 CE, before intensive forest clearance by Euro-American settlers. The surveyors partitioned the land into townships of ~36 square miles (93.2 km²), divided into 36 sections of ~1 square mile (2.6 km²). Surveyors marked two to four ‘bearing trees’ at section corners and two bearing trees at quarter-section corners midway between section corners. They recorded bearing trees in field notes by common name, diameter, and distance and bearing from the corner. The bearing-tree data for Missouri are available from the Missouri Spatial Data Information Service (http://msdis.missouri.edu). For this study, the bearing-tree distributions of Pinus, common Quercus taxa, and other trees were mapped for the entire state of Missouri to reconstruct forest composition immediately prior to Euro-American settlement (Figure 1b and c).

Coring, lithology, and chronology

Two cores, SWP-A and SWP-C, were obtained ~1 m apart with a 5-cm-diameter Wright square-rod piston corer (Wright et al., 1984) in July, 1994 CE. Cores were imaged at the National Lacustrine Core Facility at the University of Minnesota using a Geotek Geoscan-III at a 10 pixel·mm⁻¹ resolution. The two cores were stratigraphically aligned and cut into 1-cm sections to form a continuous 792-cm-long composite core.

Organic, carbonate, and non-carbonate inorganic content percent by mass values were determined by loss-on-ignition analysis (LOI; Heiri et al., 2001). Terrestrial plant remains and charcoal particles were collected for AMS radiocarbon dating during initial sub-sampling and charcoal counting. An age-depth model was developed from a series of 2 calendar ages and 10 uncalibrated AMS radiocarbon dates (Table 1) using Bacon 2.2 (Blaauw and Christen, 2011). Bacon calibrates the radiocarbon dates using Bayesian techniques that consider the stratigraphic ordering of the dates and the prior assumption that dates higher in the sequence must be younger. Consequently, the probability distribution functions of the calibrated ages can be narrower than dates calibrated individually with non-Bayesian techniques (Table 1), especially if the probability distribution functions of the non-Bayesian calibrated dates overlap, as they do from Sweeton Pond. Bacon produces weighted mean ages and 95% highest posterior density (HPD) intervals, essentially Bayesian confidence limits, for sample ages. The age model is based on the IntCal13 calibration curve (Reimer et al., 2013).

OPEN IN VIEWER

Table 1.

Calendar ages and AMS radiocarbon dates from Sweeton Pond, Missouri, US.

Age control/Lab no.	Depth (cm)	Radiocarbon age (14C yr BP)	Calibrated age (cal. yr BP; 2-σ range) a	Material dated
Core top	0–1	−44 ± 1 c
Ambrosia rise b	96–102	74 ± 3 c
UCIAMS-135272	164–165	615 ± 20	552–613, 618–653	Charcoal
UCIAMS-155806	183–184	200 ± 20	0–14, 146–189, 193–213, 268–296	Leaf fragments
UCIAMS-155807	183–184	300 ± 20	301–333, 353–435	Charcoal
UCIAMS-135273	207–208	255 ± 20	154–167, 283–316, 408–421	Charcoal
UCIAMS-155808	231–232	240 ± 20	0–3, 153–169, 282–308	Charcoal
UCIAMS-155809	232.5–233.5	240 ± 20	0–3, 153–169, 282–308	Wood twig
UCIAMS-135274	378–379	870 ± 20	730–798, 870–898	Deciduous tree leaf fragment
UCIAMS-135275	491–492	1265 ± 20	1179–1269	Brasenia seeds
UCIAMS-135276	616–617	1510 ± 20	1342–1416, 1463–1479, 1499–1515	Wood twig, partly charcoalized
CAMS-15293	786–787	1940 ± 60	1720–2004, 2028–2035	Wood

a

Calibrated with CALIB 7.10 using the IntCal13 calibration curve (Reimer et al., 2013; Stuiver et al., 2017). The input data for the Bacon age model are the raw radiocarbon dates, which Bacon calibrates using Bayesian techniques (see text).

b

Rise in Ambrosia pollen indicates local Euro-American settlement and land clearance (Holmes, 1968).

c

Dates for the core top and Ambrosia rise are calendar years before present (1950 CE).

Pollen and charcoal analyses

Samples for pollen analysis were prepared with standard procedures using a microsphere tracer and mounted in silicone oil (Faegri et al., 1989). Percentages for terrestrial taxa were calculated based on a sum of at least 300 terrestrial pollen and spores, while percentages for aquatic taxa were based on a denominator of total pollen and spores. Zonation of the pollen data was based on stratigraphically constrained incremental sum-of-squares cluster analysis (CONISS; Grimm, 1987; Grimm et al., 2011).

Samples for charcoal analysis (1 cm³) were treated with a 10% solution of tetrasodium diphosphate (Na₄P₂O₇), and then gently washed through nested 250 and 180 µm sieves. Charcoal was identified and counted at 10–40× magnification (Whitlock and Larsen, 2001). CharAnalysis Version 1.1 (Higuera et al., 2009) and REDFIT spectral analysis (Schulz and Mudelsee, 2002) were used to interpret charcoal data. Sedimentation rates were calculated from a fourth-order polynomial fit to the weighted mean ages from the Bacon age-depth model. In CharAnalysis, charcoal concentration values (particles·cm⁻³) were divided by an interpolated temporal resolution of 2 yr·sample⁻¹, the median sampling resolution of the entire record, to provide an interpolated charcoal accumulation rate (CHAR, particles·cm⁻²·yr⁻¹). A 100-yr moving median was used to describe background charcoal (BCHAR). Fire episodes were defined as the positive residual values from interpolated CHAR above the BCHAR values at the 99th percentile. To separate fire-related (signal) from non-fire related variability (noise), charcoal-peak values had to surpass this percentile of a locally defined Gaussian distribution model (Higuera et al., 2009). CHAR data, calculated separately from CharAnalysis, but also using the weighted mean values from the Bacon age-depth model, then underwent spectral analysis using REDFIT, to test for frequently re-occurring fire return intervals (FRI: years between fires) (Brown et al., 2005; Gilman et al., 1963).

Public Land Surveys (PLS) vegetation reconstruction

The bearing-tree maps show that southern and southeastern Missouri were heavily forested prior to intensive Euro-American settlement, whereas northern and western Missouri supported a mosaic of forest and prairie, with trees mainly along river valleys and grasslands generally in upland settings (Figure 1b and c). Quercus was common in forested upland areas, but not in the riparian forest along Missouri and Mississippi rivers. Pinus grew on the more rugged terrain of southeastern Missouri, west of the Mississippi floodplain. Significant disjunct populations of Pinus occurred along the North Fork River and Bryant Creek, which lie ~5 km east and west of Sweeton Pond, respectively (Figure 1c).

Individual maps for Quercus coccinea, Q. velutina, Q. palustris, and Q. rubra were grouped together (henceforth the red oak group) in Figure 1b because they showed considerable variability and misidentification by surveyors. Q. alba and the red oak group were common throughout Missouri. Q. macrocarpa occurred across the landscape in northern Missouri and in mainly riparian habitats throughout the rest of the state. However, the rectangular shape of the Q. macrocarpa distribution in northern Missouri and complete absence in extreme northeastern Missouri suggests that many surveyors misidentified it as Q. alba. Nevertheless, a geographic pattern does exist in the distribution of Quercus species favoring different growing environments. For example, Q. alba was particularly abundant in southcentral Missouri, including the area around Sweeton Pond, while Q. marilandica was most common in drier southwestern Missouri.

Chronology and lithology

Based on the Bacon age-depth model, the basal age of the core is ~1870 cal. yr BP (95% HPD interval: 2000–1750 cal. yr BP; Figure 2). Sediment was generally an organic-rich silty-clay with woody fragments and plant macrofossils throughout. Based on LOI, organic content varied from 2.8% to 41.6% of dry mass, carbonate 0.7–7.5%, and non-carbonate inorganic 57.0–96.3% (Figure 2). Deposition time changed gradually: from 3.1 yr·cm⁻¹ at 792–773 cm depth (~1870–1820 cal. yr BP), to 2.6 yr·cm⁻¹ at 773–234 cm (~1820–420 cal. yr BP), to 2.5 yr·cm⁻¹ at 234–139 cm (~420–180 cal. yr BP), and to 1.6 yr·cm⁻¹ at 139–0 cm (~180–44 cal. yr BP).

OPEN IN VIEWER

Figure 2.

Age-depth model for Sweeton Pond using Bacon 2.2 (Blaauw and Christen, 2011). Bayesian priors were as follows: acc.shape = 1.5, acc.mean = 2.5, mem.strength = 4, mem.mean = 0.7, thick = 5) Blue areas represent the probability distributions of the calibrated dates, the dotted red line shows the weighted mean age for each depth, and gray shading and dotted light gray lines depict the most likely age-depth model and 95% highest posterior density intervals. The deposition time rate (yr·cm⁻¹), percent composition of organic (light orange), carbonate (blue), non-carbonate inorganic (dark orange), and lithologic units are described in the text.

Pollen and charcoal analyses

Based on the CONISS cluster analysis, five stratigraphic zones (SWP-1 to SWP-5) were defined. The salient features are as follows (Figures 3 and 4).

OPEN IN VIEWER

Figure 3.

Pollen percentage diagram from Sweeton Pond showing dominant pollen taxa and charcoal data. Rarer taxa are provided with a 5× exaggeration (gray). Pollen zones were identified based on CONISS analysis (Grimm, 1987). Charcoal accumulation rates (CHAR, black line) and background CHAR (red line) describe variations in fire activity, and significant charcoal peaks (+) represent fire episodes. Sociocultural periods used to discuss the record are provided and explained in the text.

OPEN IN VIEWER

Figure 4.

Temperature reconstructions from Moberg et al. (2005) (orange) and Mann et al. (2008) (blue, EIV land) are plotted by z-score. The annual (gray) and smoothed (black) PDSI for the Ozark region is provided, beginning 900 CE (ca. 1050 cal. yr BP) (Cook et al., 2010). Significant peaks are plotted (red+) with their magnitudes (black) and fire return intervals (gray). Percent pollen abundance of Poaceae (orange), mesic taxa (i.e. Ulmus, Fraxinus quadrangulata-type, Juglans nigra, Morus, Ostrya/Carpinus, and Populus deltoides-type combined; blue), Pinus (green), and Ambrosia-type (pink) are overlaid and plotted with charcoal accumulation rate (black).

Zone SWP-1 (792–628 cm, ~1870–1430 cal. yr BP) is dominated by Quercus, Carya, and Pinus, with elevated Vitis, Ambrosia-type, Brasenia schreberi, and Cephalanthus at the base. CHAR is 1.3–276.5 particles·cm⁻²·yr⁻¹, with an elevated interval at 780–758 cm (~1840 to 1780 cal. yr BP). BCHAR is 7.0–95.9 particles·cm⁻²·yr⁻¹, and FRI is 27.1 yr·fire⁻¹, with 16 peaks. Peak magnitudes are 4.7–6488.3 particles·cm⁻²·peak⁻¹ (median = 176.5 particles·cm⁻²·peak⁻¹).

Zone SWP-2 (628–304 cm, ~1430–590 cal. yr BP) features increased Poaceae, Cephalanthus, and Sagittaria, with decreased Quercus. CHAR is 0.0–178.6 particles·cm⁻²·yr⁻¹; BCHAR 5.9–26.8 particles·cm⁻²·yr⁻¹; and FRI is 24.2 yr·fire⁻¹, with 39 peaks. Peak magnitudes are 1.3–4404.1 particles·cm⁻²·peak⁻¹ (median = 130.3 particles·cm⁻²·peak⁻¹).

Zone SWP-3 (304–229 depth, ~590–400 cal. yr BP) is marked by increased Salix, Nyssa, and Juglans nigra, while Poaceae, Brasenia schreberi, and Sagittaria decrease. CHAR rises to 0.4–378.3 particles·cm⁻²·yr⁻¹; BCHAR is 16.2–22.5 particles·cm⁻²·yr⁻¹; and FRI is 22.3 yr·fire⁻¹, with 13 peaks. Peak magnitudes are 1.9–9132.6 particles·cm⁻²·peak⁻¹ (median = 115 particles·cm⁻²·peak⁻¹).

Zone SWP-4 (229–102 cm, ~400–80 cal. yr BP) shows increased Pinus, Ulmus, Populus deltoides-type, Salix, and Fagus, with decreased Poaceae, Brasenia schreberi, and Sagittaria. CHAR is 2.9–414.5 particles·cm⁻²·yr⁻¹; BCHAR 15–21 particles·cm⁻²·yr⁻¹, and FRI is 16.3 yr·fire⁻¹, with 16 peaks. Peak magnitudes are 11.2–2084.7 particles·cm⁻²·peak⁻¹ (median = 184.9 particles·cm⁻²·peak⁻¹).

Zone SWP-5 (102–0 cm, ~80 to ‒44 cal. yr BP) is marked by increased Ambrosia-type with decreased Pinus, Quercus, and Cephalanthus. CHAR decreases to 0–190 particles·cm⁻²·yr⁻¹; BCHAR is 3–21 particles·cm⁻²·yr⁻¹ (mean = 6); and FRI is 22.3 yr·fire⁻¹ with six peaks. Peak magnitudes are 16.1–1237.8 particles cm⁻²·peak⁻¹ (median = 79.4 particles·cm⁻²·peak⁻¹).

REDFIT spectral analysis of the CHAR data shows that fire-episode frequencies varied throughout the record, surpassing a chi-square test with 90%, 95%, and 99% confidence intervals (CI; Figure 5). Significant FRIs (>95% CI) are 5–10 yr·fire⁻¹. Highly significant FRIs (>99% CI) are 6–8 yr·fire⁻¹. The only FRI to surpass the 99.87% false alarm CI is 7 yr·fire⁻¹.

OPEN IN VIEWER

Figure 5.

Results of REDFIT spectral analysis are plotted. CI is provided and a false alarm CI of 1-n⁻¹ – where n is the number of data points – is used in the text to highlight frequencies with the least probability of being the result of red noise (Schulz and Mudelsee, 2002; Thomson, 1990). Here, the false alarm CI is 99.87% (n = 792).

Pre-Osage period (ending ~450 cal. yr BP, 1500 CE)

From ~1840 to 450 cal. yr BP, the pollen record suggests that Sweeton Pond was surrounded by Quercus-Carya forest (Figure 3). High pollen values of the littoral shrub Cephalanthus and the aquatics Brasenia and Sagittaria in zones SWP-1 and SWP-2 suggest that the pond was initially shallow. At ~1370 cal. yr BP (95% HPD interval: 1420–1280 cal. yr BP), Poaceae pollen increased while Quercus pollen decreased but remained dominant, suggesting formation of either prairie openings, savanna, or parkland vegetation coincident with the warm and dry conditions (Cook et al., 2010; Mann et al., 2008; Moberg et al., 2005). Poaceae pollen reached its highest abundance in the record by ~660 cal. yr BP (95% HPD interval: 750–600 cal. yr BP), then abruptly declined as mesic pollen taxa (i.e. Juglans nigra, Nyssa, and Salix) increased through ~590 cal. yr BP (zone SWP-3). After ~660 cal. yr BP, Cephalanthus, Brasenia, and Sagittaria pollen decreased, suggesting that water depth increased. These changes indicate the development of more mesic vegetation and a shift to more humid conditions between ~660 and 590 cal. yr BP, coincident with a cooler and effectively wetter climate than before (Cook et al., 2010; Mann et al., 2008; Moberg et al., 2005). Following a brief period of elevated CHAR at ~1830–1670 cal. yr BP, CHAR was continuous but low from ~1670 to 450 cal. yr BP, concomitant to a decrease in FRI, suggesting frequent, low-severity fires typical of modern surface fires in open Quercus-Carya forest.

Prior to ~450 BP, human occupation of the region was likely by small, seasonal hunting parties (McMillan, 2014a; Wood and McMillan, 1976). As a result, human influence on the landscape in the southern Missouri Ozarks during this period was indistinguishable from that created by natural disturbance processes (e.g. lightning-ignited fires), in contrast with the great influence of the large Mississippian populations on the landscapes of the Mississippi and Ohio River valleys (Bird et al., 2017; Muñoz et al., 2014).

Osage period (~450–130 cal. yr BP, 1500–1820 CE)

Between ~450 and 130 cal. yr BP, pollen from mesic taxa continued to expand, particularly that of Ulmus, but also Fraxinus quadrangulata-type, Juglans nigra, Morus, Ostrya/Carpinus, and Populus deltoides-type; while Quercus, Carya, and Poaceae pollen declined. These mesic taxa likely grew in riparian habitats or on moist soils and expanded coincident with a cool and effectively wet climate. At the same time, Pinus pollen increased, suggesting that fire-dependent P. echinata expanded in the North Fork and Bryant Creek valleys on either side of Sweeton Pond, forming mixed Quercus-Carya-Pinus forest.

At Cupola Pond, which lies within the main range of Pinus in the Ozarks to the east of Sweeton Pond (Figure 1), a slight increase in Pinus at ~5000 cal. yr BP may represent the establishment of small Pinus populations in the region; however, the certain establishment of Pinus-Quercus forest in the Ozarks occurred less than 2000 years ago, when Pinus increased to about 20% (Jones et al., 2017). At Sweeton Pond, Pinus pollen abundance fluctuated between 5% and 10% until ~420 cal. yr BP (95% HPD interval: 460–370 cal. yr BP), when it reached values of 15–25%. This period of high Pinus pollen abundance is similar to the modern Pinus pollen abundance from Cupola Pond, where Pinus-Quercus forest dominates (Jones et al., 2017), and suggests local presence of Pinus (for eastern Pinus; Williams et al., 2006). Low values of Pinus pollen prior to ~420 cal. yr BP may have derived from small local populations or long-distance transport from the main Ozark range to the east. The subsequent increase in Pinus pollen by ~420 cal. yr BP indicates certain establishment and expansion of P. echinata near Sweeton Pond.

Increased CHAR from ~410 to 130 cal. yr BP suggests increased fire activity occurred in conjunction with the development of mixed Quercus-Carya-Pinus forest, and with a cooler and effectively wetter climate than before. Increased fire activity despite effectively wetter conditions could have resulted from greater fuel biomass or increased ignition. Although fuel loads and fire activity in shortgrass prairie are positively correlated with moisture (Grimm, 2011; Nelson et al., 2004), this relationship may not hold for higher biomass vegetation types, such as tall-grass prairie and savanna (Nelson et al., 2006). Indeed, within the forest but near the prairie-forest border in northwestern Minnesota, Clark (1988a, 1990) documented decreased fire frequency with increased humidity over the past 750 years, suggesting that increased fire ignition was an important driver of increased fire activity near Sweeton Pond after ~410 cal. yr BP.

Notably, fire activity did not increase with the initial expansion of mesic species between ~750 and 590 cal. yr BP, but contemporaneous with the arrival and subsequent expansion of the Osage 150–200 years later. Following French contact in the region at ~1680 CE (270 BP), the Osage expanded their territory and increased the scale of their hunting and warring to provide game and slaves in exchange for French goods, likely utilizing fire in the process (Ponziglione, 1882; Schoolcraft, 1821, 1853).

Given that this region has few lightning-ignited fires (<1 per 4000 km² annually, accounting for ~1% of fire occurrences in recent decades) (Guyette et al., 2002; Schroeder and Buck, 1970; Stambaugh and Guyette, 2008) and that the arrival of the Osage coincided with increased CHAR and Pinus pollen, we propose that the combination of increased humidity and fire ignition by the Osage promoted the expansion of P. echinata in the North Fork and Bryant Creek valleys. The expansion of both fire-sensitive and fire-dependent tree species coincident with Osage occupation of the area suggest that anthropogenic fire and land use were local in nature and increased landscape heterogeneity prior to Euro-American settlement. This hypothesis aligns with that of Stambaugh and Guyette (2008), who documented decreasing fire activity with distance from the rivers, which were foci for human habitation and travel in the Current River watershed of southeastern Missouri, between 1620 and 1850 CE (330–170 BP). Furthermore, the hypothesis that Osage fire and land use increased landscape heterogeneity in the Ozarks supports the argument that Native American land use altered vegetation and fire history to varying degrees and at varying spatial scales in the region, prior to Euro-American settlement (Gajewski et al., 2017; Munoz et al., 2014; Vale, 2002).

Euro-American period (since ~130 cal. yr BP, 1820 CE)

The forced removal of the Osage from the Missouri Ozarks by 1820 CE, and increased Euro-American settlement of the region preceded dramatic changes in the local vegetation and fire history that only intensified with later population growth following the end of the American Civil War (1861–1865 CE) (Figure 3; Bailey, 2001; Holmes, 1968). After ~1850 CE (100 cal. yr BP; 95% HPD interval: 140–80 cal. yr BP), Ambrosia-type pollen increased dramatically at the expense of Pinus, Quercus, and Carya pollen. By 1880 CE, the timber and mining industries spurred Euro-American immigration to the post-war Missouri Ozarks and most of the P. echinata in Ozark county (immediately south of Sweeton Pond) was removed by logging (Holmes, 1968). This is clearly represented by a sharp decline in Pinus pollen between ~1850 and 1890 CE (100–60 cal. yr BP; ending 95% HPD interval: 80–40 cal. yr BP) from 24.8% to 2.4%. Charcoal is nearly absent after ~1890 CE (60 cal. yr BP; 95% HPD interval: 70–40 cal. yr BP), likely pre-dating the active suppression of wildfires that began in the 1920s CE. Early absence of fire on the landscape probably relates to fuel fragmentation (Cutter and Guyette, 1994; Guyette et al., 2002; Guyette and McGinnes, 1982). Since the beginning of fire suppression efforts, fire-intolerant broadleaf species have been replacing Quercus-Carya-Pinus forest throughout the Missouri Ozarks (Hanberry et al., 2014a, 2014c).

In addition, row-crop farming for wheat and corn led to increased soil runoff in the Sweeton Pond catchment, which is visible after ~1880 CE (70 cal. yr BP; 95% HPD interval: 80–60 cal. yr BP), when deposition time and non-carbonate inorganic matter increased (Figure 2). As erosive row-crop farming was replaced first by pasturing and then by zero-till practices, organic content rebounded and non-carbonate inorganic content decreased to pre-Euro-American settlement levels.

Summary of fire severity and return interval

Despite the variation in vegetation and the magnitude of charcoal peaks throughout the record, the CharAnalysis-reconstructed FRI remained relatively invariant, between 22 and 28 years·fire⁻¹, which suggests that variation in fuel load was not the primary factor driving fire-episode frequency, as it was, for example, in northwestern Minnesota, where FRI varied as much as 35 years·fire⁻¹ with changes in vegetation and climate (Clark, 1988b, 1990). REDFIT spectral analysis of the CHAR data calculated independently of CharAnalysis showed that FRI was close to 5–10 yr·fire⁻¹ (Figure 5). The 5–10 yr·fire⁻¹ FRI from REDFIT is supported by nearby fire-scar records, which suggest a FRI as low as 3–4 yr·fire⁻¹ prior to Euro-American settlement (Cutter and Guyette, 1994; Guyette and McGinnes, 1982). The 5–10 yr periodicity in FRI recorded at Sweeton Pond falls within the ENSO band (Moy et al., 2002), which suggests that ENSO may have been a driving factor for moisture variability for the past 1900 years in the Ozark region. This conclusion is consistent with the findings of Guyette et al. (2006) from 100 km south in the Boston Mountains of Arkansas, USA.

Charcoal-peak magnitude can be affected by fire episode size, severity, proximity to the catchment area, and taphonomic processes (Clark et al., 1998; Higuera et al., 2007; Walsh et al., 2010; Whitlock and Millspaugh, 1996). Given the overestimation of FRI by CharAnalysis as compared with that from REDFIT spectral analysis and nearby fire-scar records, we assume that the charcoal peaks identified by CharAnalysis were produced by fire episodes in close proximity to Sweeton Pond. The local source area (<3 km radius from the site) of charcoal peaks has been noted in other studies (Higuera et al., 2010; Kelly et al., 2011). Furthermore, we doubt that erosion played a significant role in peak magnitude because non-carbonate inorganic content varies little until the establishment of row cropping ~1880 CE, suggesting that the influx of soil surrounding the pond was minimal and that any re-deposition of charcoal particles was likely registered as BCHAR. We therefore interpret high peak magnitudes as fire episodes of greater size or severity, as compared with fire episodes with of low peak magnitude (Higuera et al., 2014).

Median charcoal-peak magnitude was highest in zones SWP-1 (1870–1430 cal. yr BP) and SWP-4 (400–80 cal. yr BP), indicating greater size or severity of fire episodes during these times (Figure 4). In zones SWP-2 and SWP-3 (1430–400 cal. yr BP), median peak magnitude is lower, but these zones still have 8 of the 14 largest charcoal peaks (>1000 particles·cm⁻²·peak⁻¹). Zones SWP-2 and SWP-3 coincide with a period of increased Northern Hemisphere temperature from ~1050 to 550 BP (Mann et al., 2008, 2009; Moberg et al., 2005), encompassing the MCA. Increased Poaceae pollen at the expense of arboreal taxa in zones SWP-2 and SWP-3, higher Northern Hemisphere temperatures, and low PDSI values suggest that this period was warm with effectively dry summers. The occurrence of high individual charcoal peaks amid generally lower CHAR and median peak magnitudes suggests that surface fires were punctuated by large or severe fire episodes between ~1430 and 400 cal. yr BP.

The highest peak in the record (9132.6 particles·cm⁻²·peak⁻¹) at ~410 cal. yr BP (95% HPD interval: 450–300 cal. yr BP) occurred in zone SWP-3 and was followed by increased BCHAR during zone SWP-4, suggesting that fire activity increased between ~410 and 80 cal. yr BP. Somewhat paradoxically, increased abundance of mesic arboreal taxa and low reconstructed temperatures suggest that this period was cooler and wetter than before, coincident with LIA (Mann et al., 2008, 2009; Moberg et al., 2005). High BCHAR in zone SWP-4 between ~410 and 80 cal. yr BP and decreased FRIs reconstructed from fire scars during the same time (Guyette et al., 2002) suggest that increased fire activity was likely as a result of increased fire ignition. Given the extremely low modern occurrence of lightning-ignited fires in the region, we ascribe the increase in fire activity during this period first to the Osage, who extensively occupied the area between ~1550 and 1820 CE (400–130 BP) (Bailey, 1973, 2001), and then later to Euro-American land clearance.