Compiling historical descriptions of past Indigenous cultural burning: a dataset for the eastern United States

Introduction

There remains a debate over the geographic extent and intensity of past Indigenous land use, including fire, upon historical ecosystems of the eastern United States. Past Indigenous cultures employed various land-use practices that included cultural burning for reasons such as to prepare land for agriculture, clear landscapes for easing travel, promote edible or other favoured fire-tolerant plant species, create open woodlands with grasses that nourished game animals, and facilitate the hunting of game (Day 1953; Williams 2000; Stewart 2002; Abrams and Nowacki 2008). The Indigenous population in 1500 totalled between 0.5 and 2.6 million in the eastern US, though estimates vary widely (Milner and Chaplin 2010). Scholars have continued to renew this debate over several decades, most recently by paleoecologists and archaeologists claiming that Indigenous groups minimally impacted fire regimes over millennia in southern New England (Oswald et al. 2020), a claim that other researchers (Abrams and Nowacki 2020; Roos 2020) and Indigenous persons (Leonard et al. 2020) dispute. This exchange was only the most recent iteration of the debate (Denevan 1992, 2011; Vale 2002; Mann 2006), one extending back throughout the 20th century (e.g. see summaries in Patterson III and Sassaman 1988; Whitney 1996).

A related debate concerns the causes of ‘mesophication’ of forests during the 20th and 21st centuries, in which shade-intolerant, fire-tolerant tree species typical of open forests (e.g. oak [Quercus spp.] and pine [Pinus spp.]) are being replaced by shade-tolerant, fire-sensitive species (e.g. maple [Acer spp.]) to form closed forests (Nowacki and Abrams 2008). Researchers implicated various causes in this transition, including fire exclusion, climatic changes, changes in herbivore populations, and modern land-use change (McEwan et al. 2011; Hanberry et al. 2020), but fire exclusion is often argued as the main driver of mesophication. Understanding the extent and intensity of land clearance and burning is of utmost importance when ascribing potential causes of mesophication, and when setting modern land management targets. Understanding the extent of past burning influences discussions over whether to maintain declining fire-dependent plant communities and open forests (e.g. Ryan et al. 2013; Abrams et al. 2022), or to maximise carbon sequestration by protecting or promoting old-growth closed forests (e.g. Oswald et al. 2020; Kellett et al. 2023), goals that can directly oppose each other.

Researchers have examined descriptions from early explorers, colonists, travellers, naturalists, and historians of the circa 17th–19th centuries CE, to understand the spatial and temporal dimensions of Indigenous burning (Table 1). Notable syntheses of these descriptions include Flannery (1939) and Day (1953) for the northeastern US, Whitney (1996) for the northeastern and midwestern US, Cronon (1983) for New England, Higgins (1986) for the Northern Great Plains, McClain et al. (2021) for the Midwestern Tallgrass Prairie, Stewart (2002) for the Lower 48 US, Pyne (1982) and Williams (2005) for North America, and Scherjon et al. (2015) globally. Descriptions can provide eyewitness accounts of Indigenous peoples as an ignition source (as opposed to lightning), establish relationships between burning and open forest or grassland conditions, describe densification of forests once Indigenous burning ceased, and provide evidence of the seasonality of burning (e.g. Higgins 1986; McClain et al. 2021). While not all descriptions directly observed burning, some describe open forest conditions dominated by fire-adapted vegetation and attribute it to Indigenous land management. Early descriptions of burning have also been incorporated into land management plans (e.g. Keister 1998) and landscape histories (Cronon 1983; Tyndall and Hull 1999).

Authors of such syntheses of historical documents generally conclude that in the eastern US, forest and grassland burning was geographically widespread, was more concentrated in areas near Indigenous settlement, and occurred in climates that supported burning (e.g. Pyne 1982; Stewart 2002). While some have questioned the motivations of early observers (Russell 1983; Russell and Forman 1984), others have pointed out that respected early naturalists linked fire to large open landscapes when traveling eastern North America (Whitney 1996). For example, along the Ohio River circa 1788, a scientist and historian, Hildreth (1848), described open forests stating that Indigenous burning each autumn ‘destroyed all the shrubs and undergrowth of woody plants.’ A few studies have compared historical descriptions to other data or methodologies and found that descriptions correspond geographically with fire-adapted vegetation, woodlands, and/or grasslands described in historical land surveys, and/or model predictions of fire-dependent vegetation communities (Black et al. 2006; Tulowiecki et al. 2020a). Descriptions from historical documents complement and contrast with other evidence of burning from pollen or charcoal in paleoecological records (e.g. Oswald et al. 2020), tree-ring fire-scar records (e.g. Marschall et al. 2022), early land survey records (Thomas-Van Gundy et al. 2015; Tulowiecki et al. 2020a), and ethnographies (e.g. Parker 1910; Ruffner 2005).

Despite the value of historical descriptions to environmental historians and biogeographers for qualitatively assessing characteristics of Indigenous fire use, and to physical scientists as a form of validation for other methods, few efforts have sought to discover, catalogue and map early descriptions of Indigenous burning to make them more available for study and interpretation. Higgins (1986) assembled a compendium of nearly 300 firsthand historical accounts of burning in the Northern Great Plains from 1673 to 1880. Williams (2005) compiled an annotated bibliography predominantly of scholarly texts, but also some 17th–19th century historical descriptions on past Indigenous burning in North America. Scherjon et al. (2015) created a dataset of 231 descriptions of Indigenous burning from early ethnographies and historical texts, including 87 for North America. Using historical texts in western New York, Tulowiecki et al. (2020a) used information retrieval methods to find descriptions of Indigenous burning by searching for paragraphs containing keywords related to fire and Indigenous peoples. This lattermost study discovered 13 locations with descriptions in their study area and the descriptions were often located within 15 km from former Indigenous villages and trails. McClain et al. (2021) compiled 795 historical mentions of fire in the Midwestern Tallgrass Prairie, of which 32 were attributed to Native Americans between 1673 and 1836. Comparable efforts have mapped or compiled descriptions of past Indigenous agricultural techniques contained within early historical accounts and ethnographies across North America (Flannery 1939; Doolittle 2002). Worth noting are general searchable databases of the accounts of explorers and related figures in early American history that mentioned fire (e.g. Alexander Street 2024; Wisconsin Historical Society 2024).

This paper summarises an ongoing effort to compile and map historical descriptions of cultural burning by past Indigenous cultures in the eastern US. It presents an initial work for the northeastern US. Included in this compendium are previously cited descriptions both from earlier research, and descriptions discovered via information retrieval methods. It shares accounts as a GIS-format layer, a tabular file, and an interactive web map for those studying the distribution and characteristics of burning. It is expected that sharing this dataset will provide means of scholarly appraisal of such descriptions, and aid those seeking to compare and contrast historical descriptions with other data and methodologies that assess past fire regimes.

Compiling descriptions

This study first gathered historical descriptions of Indigenous burning from previously published scholarly works. These works include Bromley (1935), Day (1953), Stewart (2002), Pyne (1982), Cronon (1983), Whitney (1996), Williams (2005), and Scherjon et al. (2015). Also included were descriptions from publications specific to western New York (Tulowiecki et al. 2020a, 2020b). At the time of publication of this article this study focused upon gathering historical descriptions of the nine-state region of the northeastern US: Pennsylvania, New Jersey, New York, Connecticut, Rhode Island, Massachusetts, Vermont, New Hampshire, and Maine.

Also, a previously published information retrieval method (Tulowiecki et al. 2020b) was used for discovering new descriptions within digitised historical texts to augment the collection of descriptions derived from previous syntheses. Over 400 plain text format (.txt) county histories were searched, which were assembled from the author’s previous publication (Tulowiecki 2018). Thousands of digitised historical texts were also downloaded from the Internet Archive (2023), spanning the 16th–20th centuries that met keyword searches for terms such as ‘history’ along with state names. The ‘Cygwin’ terminal emulator and the ‘wget’ package (Chamberlain et al. 2022) was utilised in this latter approach to download these additional plain text documents in bulk, following a tutorial on the Internet Archive (Kaplan 2012). The major types of texts obtained from this lattermost approach were county, town, and other histories. An R script (R Development Core Team 2022; RStudio Team 2022) then broke plain text documents into paragraphs, tallied the number of certain keywords (e.g. ‘Indian’, ‘fire’, etc.) per paragraph, applied a predictive model to assign the paragraph a score based on keyword frequencies, and ranked the paragraphs by score in the final output (Tulowiecki et al. 2020b).

Summarising descriptions

Once descriptions were collected, several attributes were recorded in the initial version of the dataset (Table 2). Regarding the description itself, the following attributes were recorded: a one-sentence excerpt, a locational description, estimated latitude and longitude, an estimate of its positional uncertainty, description type (e.g. firsthand observation, historical summary), and the name of the firsthand observer if applicable. Also, recorded was an estimated associated year: either the year the burning occurred as stated by the observer, the estimated date the landscape was last burned, or the estimated date of the first white settlement, as provided within, preceding, or following the description. Metadata on the document containing the description were recorded such as author, title, year published, publisher, place published, and document type (e.g. county history, early journal/letter/account). To allow others to find and read the full description within the document, three additional attributes were recorded: the starting page number of the description, the first sentence in the paragraph of the description, and a hyperlink to the page containing the description within a scanned PDF on Internet Archive or Google Books. If present, also recorded were place names (toponyms) related to the burning or fire-related landscape conditions associated with the description, and their language of origin. Finally, occasional notes from the author of this publication were provided concerning various issues such as ambiguities interpreting the description. Unlike previous work that developed a dataset from more formal ethnographies (Scherjon et al. 2015), this study did not record attributes related to purposes of burning because in comparison to previous work, the descriptions analysed here were less detailed regarding characteristics of fire use.

Descriptions were then mapped as points representing the ‘best-guess’ location where the burning occurred, typically the centre of the geographic area associated with the description. Points versus polygons were utilised because they allowed greater cartographic clarity and flexibility in portraying the descriptions (see next section). Points describing the same geographic area were plotted on top of each other. A portion of descriptions was interpreted and digitised by students in upper-level undergraduate geographic information systems (GIS) courses.

Sharing descriptions

Currently the descriptions are provided as a GIS point vector layer (geodatabase [.gdb] feature class), a tabular file (.xlsx), and an interactive web map (https://go.geneseo.edu/HDICB). The decision to provide the data as a single GIS layer, rather than as a relational database (e.g. with document metadata in one table, descriptions in another, etc.), was made to provide ease of access for potential incorporation into future studies and comparison with findings from other methods of studying past fire regimes in GIS software. The GIS layer and tabular file versions are available from the NOAA NCEI World Data Service for Paleoclimatology archives at https://www.ncei.noaa.gov/access/paleo-search/study/39659.

The interactive ArcGIS Story Map currently contains multiple thematic maps (Table 3) that convey various aspects of the descriptions based upon their attributes (Table 2). Each thematic map includes clickable symbols with pop-ups that show the attributes described above, along with a hyperlink to the full description. Where multiple descriptions existed, a pop-up yields multiple descriptions that users can page between. Sometimes utilised were proportional symbol pie chart symbology: the size of the point symbol conveys the number of descriptions within an area, and the size of each ‘slice’ represents a percentage of total descriptions that possessed a certain attribute (Fig. 1). Often utilised was automated point clustering: the number of descriptions aggregated into a common proportional symbol is dependent upon the scale at which a map user views the map, yielding a cleaner presentation of symbology. Also, a call to historians and scholars to share other known descriptions contained within texts will be included in the Story Map.

Fig. 1.

A screenshot of a proportional-symbols pie chart map within the current web map (https://go.geneseo.edu/HDICB), showing the number and type of descriptions in a general location. Larger circles represent a higher number of descriptions.

Descriptions compiled for the northeastern US

Results presented here summarise descriptions from the northeastern US compiled at the time of publication. A total of 253 unique combinations of description and location were mapped throughout the nine-state region (Fig. 1). Some descriptions described more than one location: for example, one traveller speculated on the causes of ‘oak plains’ in western New York at 11 locations (Dwight 1828). Around 200 geographic locations (e.g. a town, a county, a region) possessed at least one description. Around half of the descriptions described burning somewhere within towns, comparably-sized municipal divisions (e.g. boroughs, townships), or smaller municipal divisions (e.g. villages). Around one quarter described burning or burned landscapes at specific locales (e.g. hills, islands, locations along rivers), and around one tenth described burning or burned landscapes somewhere within a county. The remaining descriptions pertained to burning across or somewhere within states or regions of variable sizes. About two thirds of the descriptions possessed ‘low’ positional uncertainty, resolvable to somewhere within an approximately 100 km² area. Just over 70% of descriptions were summaries from histories (often 19th century), whereas around 10% were firsthand observations of burning. Just 7 out of 253 were historical descriptions of specific burning events, whereas 20 were contemporary observers of early burning who summarised their observations. Around one quarter of descriptions mentioned place names associated with fire or burned landscapes, in either English (e.g. ‘Burnt Hills’) or Indigenous languages. Year associated with the fire event, burned landscape, or date of white settlement ranged from 1524 to 1867 (interquartile range = 1643–1783) with a median of 1675, and manifested a clear pattern from east (older dates) to west (more recent dates). Dates for firsthand observations ranged from 1524 to 1867 (interquartile range = 1643–1804) with a median of 1795. Note that the date 1524 was from John de Verrazzano’s ambiguous description of a ‘well peopled’ region with a ‘great number of fires’ around present-day Narragansett Bay, Rhode Island (1860). Nearly 180 documents provided the historical descriptions mapped in the northeastern US; the middle 50% (interquartile range) of publication dates ranged from 1861 to 1898, with a median date of 1883 and the earliest date of 1634.

The current dataset is representative of major themes in previous research such as the connections between fire and open forests, woodlands, or savannas with fire-tolerant species (Fig. 2) such as oak, hickory (Carya spp.), pine, and chestnut (Castanea dentata); reforestation after fire exclusion; the use of fire predominantly for clearing landscapes for agriculture, hunting, and travel benefits; and continuity in burning from Indigenous to initial white possession of the land. Descriptions in the current dataset also contain less-emphasised themes from previous research. While most descriptions state that the purpose of burning was to clear landscapes, occasional descriptions mention the use of fire to drive away enemies during conflicts, and at least three descriptions (two purportedly firsthand) in western New York and northwestern Pennsylvania mentioned the use of burning to kill rattlesnakes (Crotalus spp.; Sistrurus spp.). Occasionally, writers or observers estimated the sizes of burned areas: 3 km² on Cape Cod in 1603 (Pring 1906) and 18 km² in the Genesee Valley in 1795 (Turner 1849) in single events; 8 km in length near Plymouth from a contemporary observer (Winslow 1622); and 4 km² in the Genesee Valley (Clayton 1879) or even 526 km² in southeastern Pennsylvania (Rupp 1845) in general descriptions of fire-maintained areas. Spatial correspondence between place names and other types of descriptions existed: for example, firsthand accounts of burning that occurred on ‘Groveland Hill’ in western New York, and ‘Schooset’ or ‘fire place’ (Huden 1962) and ‘Ragged Plain’ (Wood 1865) coincided with an area with forest regrowth after fire exclusion in 17th century Plymouth County, Massachusetts. Descriptions also occasionally speculate on the frequency and seasonality of burning, often describing burning as occurring once or twice per year in the spring and/or autumn. However, the seasonality suggested from the few firsthand observations of specific events that possessed dates was variable: early June 1769 in central New York (Smith 1906), early September 1609 off the coast of New Jersey (Juet 1860), and early December 1632 off the Delaware Coast (de Vries 1853), for example.

Fig. 2.

Historical descriptions of Indigenous burning from this study versus estimated pre-Euro-American fire-adapted tree taxa, namely oak, hickory, chestnut, and pine combined. Fire-adapted tree taxa abundance was statistically estimated from witness-tree tallies of original land survey records circa 1620–1825 (Paciorek et al. 2016). Some descriptions have been slightly moved so that all are visible. Only descriptions with ‘low’ or ‘medium’ positional uncertainty are shown (n = 217).

A few descriptions incorporated legal anecdotes related to Indigenous burning. For example, at least eight towns - but ‘probably most, if not all the towns’ (Wood 1828) - passed laws requiring residents to remove underbrush that established after fire exclusion in the plains of western Long Island; the same towns created laws to conserve timber by preventing the sale of timber outside the town. In 1671 the New York governor even required brush removal across the entire island due to deteriorating grasslands from invading shrubs and saplings – perhaps one of the earliest references to mesophication and forest densification in the eastern US. Early towns in the Connecticut River Valley also enacted brush-clearing and wood conservation laws soon after Euro-American settlement (Judd 1860). In Mastic on eastern Long Island in 1700, one description noted an early white landowner who leased lands to 11 Indigenous persons to plant crops and burn the forest (Bayles 1882).

From these initial accounts mapped, four locations emerged as major regions with altered landscapes through Indigenous burning, judging from the number and veracity of the accounts: southern New England, western Long Island, southeastern Pennsylvania, and western New York (Fig. 1). Southern New England contained multiple canonical firsthand accounts of Indigenous burning in the 17th century (Winslow 1622; Williams 1643; Wood 1865; Morton 1883), fire-related place names, and descriptions of early white settlers who appropriated Indigenous-maintained fields and hunting grounds (e.g. within the Connecticut River Valley). As noted above, writers in western Long Island described large treeless tracts that whites adopted as early pastures in the 17th century. Writers in southeastern Pennsylvania described expanses maintained in a treeless condition by Indigenous burning until around 1731, after which time forest regrowth was noted (Table 1). Relying on previous compilations of historical descriptions (Tulowiecki et al. 2020a, 2020b), western New York contained numerous descriptions of oak savannas and burned landscapes including a few firsthand accounts centred upon the Genesee River Valley and western Finger Lakes in the late 18th century.

Correspondence with previous scholarship

This study produced a comparable number of historical descriptions or accounts as previous similar compendiums (e.g. Higgins 1986; Scherjon et al. 2015). Unlike previous efforts, however, this study spatialised historical descriptions to produce new geographic insight. Initial mapping of historical descriptions generally supported the conclusions of previous scholars utilising descriptions to reveal spatial and temporal characteristics of Indigenous burning (e.g. Pyne 1982; Stewart 2002): it was geographically widespread in regions with both Indigenous settlement and climates conducive to fire. While the current dataset indicated that relatively few descriptions were firsthand accounts, and thus supports previous statements regarding the paucity of such accounts (Russell 1983), descriptions by 19th century historians collectively suggest the plausibility of local- to regional-extent influences of fire upon forested ecosystems in portions of the northeastern US (Fig. 1), and refutes research claiming minimal influence of Indigenous groups upon fire regimes and forested ecosystems in the northeastern US (Oswald et al. 2020). Historical descriptions generally corresponded with the distribution of fire-adapted tree taxa (i.e. oak, hickory, chestnut, and pine) prior to Euro-American settlement (Fig. 2), especially in southern New England, suggesting at least a partial causal relationship between fire and fire-adapted vegetation in addition to climatic and soil factors (Thomas-Van Gundy et al. 2015; Tulowiecki et al. 2022). The initial collection of descriptions also reinforces and suggests the temporal sequence of changing fire regimes: it reveals a clear east-to-west land-use transition with the exclusion of Indigenous fire, and replacement with early Euro-American burning practices (Pyne 1982; Stambaugh et al. 2018; Margolis et al. 2022).

Data quality considerations

This paper acknowledges data quality issues when interpreting the content and patterns in descriptions of Indigenous burning. The geographical pattern in historical descriptions is not an unbiased sample: descriptions were generally produced where little or no time elapsed between Indigenous and white settlement, and where Euro-American population density was high enough to produce individuals that wrote historical documents. The earliest descriptions of burned landscapes from colonists, and particularly those who also describe easily-traversed open landscapes, may be painting overly favourable views of the New World to entice settlers to emigrate (Russell 1983; Russell and Forman 1984; Patterson III and Sassaman 1988). Descriptions are often the perspectives of avocational historians, white settlers, travellers, missionaries, trappers, and others, oftentimes without scholarly backgrounds. Some may be speculative, biased, be highly derivative of earlier texts, possess high positional uncertainty, or mischaracterise the intent of Indigenous use of fire. Like previous research (Higgins 1986), in a few descriptions it was unclear whether Indigenous or early whites were burning forests: for example, a traveller in 1803 noted mountains ‘for more than fifty miles…burning…by the hunters’ (Harris 1805).

On the other hand, some 19th century historians purportedly performed rigorous analysis of early town documents and firsthand accounts to reach conclusions regarding the geographical extent of Indigenous burning (e.g. Wood 1828; Judd 1860). Moreover, previously mentioned studies have corroborated historical descriptions using other evidence such as early land survey records (e.g. Tulowiecki et al. 2020a), giving greater weight to such descriptions. Historical descriptions of burning compiled in this study thereby possess varying levels of reliability across four centuries, requiring caution during future interpretation.

Conclusion and future directions

This paper presented a novel and ongoing effort to develop and disseminate a dataset of historical descriptions of Indigenous cultural burning from historical texts. Future work will expand upon this dataset in later version releases by adding new descriptions from throughout the eastern US with comparable document density, coding new attributes, and calling for historians and scholars to contribute additional descriptions from undigitised and/or obscure documents. This effort could also be expanded by incorporating Indigenous knowledge and recollections in public memory. Coupled with previous efforts (e.g. Stewart 2002; Williams 2005; Scherjon et al. 2015; Tulowiecki et al. 2020b; McClain et al. 2021), this paper demonstrates that Native American cultural burning was described at least hundreds to thousands of times in historical documents in North America. Such historical descriptions, and methods used in this study to discover them, may also be of value to understanding historical distributions of fauna and their connections to semi-open ecosystems and fire regimes. This study also suggested a few new directions in research; for example, a greater analysis of early colonial policies (Kawashima and Tone 1983) related to brush-clearing and timber conservation may suggest locations where Indigenous burning maintained semi-open landscape conditions that early Euro-Americans encountered. The dataset developed here represents only one line of evidence regarding the geographic extent and intensity of Indigenous burning, and it is acknowledged that it possesses biases and other forms of limitations. Echoing previous calls for multiproxy evidence at local scales (Scherjon et al. 2015), this resource can contribute to future multidisciplinary or interdisciplinary research that compares and contrasts historical descriptions with findings from other methodologies, to lend greater confidence in the reconstruction and assessment of past fire regimes.