Journal of Biogeography, 29, 1279–1304 The forests of presettlement New England, USA: spatial and compositional patterns based on town proprietor surveys Charles V. Cogbill1*, John Burk2 and G. Motzkin2182 Walker Lane, Plainfield, VT, USA and 2Harvard Forest, Harvard University, Petersham, MA, USA Abstract Aim This study uses the combination of presettlement tree surveys and spatial analysis to produce an empirical reconstruction of tree species abundance and vegetation units at different scales in the original landscape. Location The New England study area extends across eight physiographic sections, from the Appalachian Mountains to the Atlantic Coastal Plain. The data are drawn from 389 original towns in what are now seven states in the north-eastern United States. These towns have early land division records which document the witness trees growing in the town before European settlement (c. seventeenth to eighteenth century AD). Methods Records of witness trees from presettlement surveys were collated from towns throughout the study area (1.3 · 105 km2). Tree abundance was averaged over town- wide samples of multiple forest types, integrating proportions of taxa at a local scale (102 km2). These data were summarized into genus groups over the sample towns, which were then mapped [geographical information system (GIS)], classified (Cluster Analysis) and ordinated [detrended correspondence analysis (DCA)]. Modern climatic and topo- graphic variables were also derived from GIS analyses for each town and all town attributes were quantitatively compared. Distributions of both individual species and vegetation units were analysed and displayed for spatial analysis of vegetation structure. Results The tally of 153,932 individual tree citations show a dominant latitudinal trend in the vegetation. Spatial patterns are concisely displayed as pie charts of genus com- position arrayed on sampled towns. Detailed interpolated frequency surfaces show spatial patterns of range and abundance of the dominant taxa. Oak, spruce, hickory and chestnut reach distinctive range limits within the study area. Eight vegetation clusters are distinguished. The northern vegetation is a continuous geographical sequence typified by beech while the southern vegetation is an amorphous group typified by oak. Main conclusions The wealth of information recorded in the New England town pre- settlement surveys is an ideal data base to elucidate the natural patterns of vegetation over an extensive spatial area. The timing, town-wide scale, expansive coverage, quantitative enumeration and unbiased estimates are critical advantages of proprietor lotting surveys in determining original tree distributions. This historical–geographical approach produces a vivid reconstruction of the natural vegetation and species distri- butions as portrayed on maps. The spatial, vegetational and environmental patterns all demonstrate a distinct Ôtension zoneÕ separating Ônorthern hardwoodÕ and Ôcentral hardwoodÕ towns. The presettlement northern hardwood forests, absolutely dominated by beech, forms a continuum responding to a complex climatic gradient of altitude and latitude. The oak forests to the south are distinguished by non-zonal units, probably affected by fire. Although at the continental scale, the forests seem to be a broad transition, at a finer scale they respond to topography such as the major valleys or the northern mountains. This study resets some preconceptions about the original forest, such as the overestimation of the role of pine, hemlock and chestnut and the *Correspondence: 82 Walker Lane, Plainfield, VT 05667 USA. E-mail: [email protected] Ó 2002 Blackwell Science Ltd 1280 C. V. Cogbill et al. underestimation of the distinctiveness of the tension zone. Most importantly, the forests of the past and their empirical description provide a basis for many ecological, educational and management applications today. Keywords Historical ecology, New England, northern hardwood forest, plant biogeography, presettlement vegetation, proprietory town, surveyor’s records, witness tree, vegetation classification, tension zone. common surrogates of past forests have questionable appli- INTRODUCTION cability in quantifying forest vegetation before settlement. Starting in 1620, settlers from Europe profoundly changed Altogether inferential methods, such as modern veget- the supposedly ÔinfiniteÕ woods of New England (Cronon, ation, ecological models, and ⁄ or theoretical relationships to 1983). By 1850, the land in the current states of Connecti- environment are problematic in determining a spatially cut, Rhode Island, Massachusetts, Vermont and New comprehensive and temporally accurate view of the historic Hampshire had been completely settled and more than 60% forest of New England. There are, however, empirical of the land over the entire region cleared for agriculture observations which describe the forests of the time. These (Harper, 1918). Areas remaining in forest, mostly on shal- historical data can also be used to test the accuracy of sur- low soils, steeper slopes or low-productivity land, were being rogate inferences. Contemporary observations of explorers, relentlessly harvested for timber by settlers and businessmen naturalists, diarists, authors and publicists abound, although (Williams, 1989). Thus by the middle of the twentieth cen- they are subjective, limited in coverage and typically qual- tury, virtually the entire forest had been altered by human itative (Whitney, 1994; Russell, 1997; Bonnicksen, 2000; activities (Whitney, 1994). Edmonds, 2001). Significantly, the classical syntheses of the Today less than a small fraction of 1% of the forest in the native vegetation across New England (e.g. Nichols, 1913; north-eastern United States remains as a few fragmented Bromley, 1935; Cline & Spurr, 1942; Braun, 1950; Westveld scraps of Ôold-growthÕ landscape (Davis, 1996). Although et al., 1956; Bormann & Buell, 1964) are based on a com- reconstruction of the nature of the presettlement landscape is bination of anecdotal accounts, together with sampling in severely hampered by the lack of modern analogues, several the last putative remnants and intensive field knowledge of approaches can be used to infer the nature of historical the existing vegetation. Interestingly, despite explicitly link- forests by extrapolating from modern data (Whitney, 1994; ing these proverbial inferences (variously called ÔnaturalÕ, Russell, 1997; Egan et al., 2001). For example, much of ÔvirginÕ, ÔclimaxÕ, ÔoriginalÕ, ÔprimevalÕ or Ôold-growthÕ forests) New England is (re)forested today and it has been posited to the past, none of these studies compare their theorized that the ÔrecoveredÕ vegetation appears similar to the forests species composition or pattern of forest types with actual before 1775 (MacCleery, 1992). Moreover, several proxies historical data. of the ostensibly original forest have been drawn from dif- Fortunately, a spatially comprehensive and temporally ferent sources: ÔvirginÕ remnants (e.g. Nichols, 1913; Cline & relevant representation of past vegetation is contained in the Spurr, 1942; Braun, 1950); successional tendencies and sil- land division surveys carried out in anticipation of European vicultural experience in managed stands (e.g. Hawley & settlement. In the Midwest, the United States General Land Hawes, 1912; Westveld et al., 1956); and theoretical models Office (GLO) surveys have long been the primary resource for of forest development and response to the environment, hundreds of studies of the historical landscape, and several particularly climate (e.g. Weaver & Clements, 1938; Bormann states have recently digitized their entire survey database & Likens, 1979; Pacala et al., 1993). Yet each of these (Whitney, 1994; Whitney & DeCant, 2001). These federal approaches has its limitations. ÔRemnantÕ stands, by defini- surveys typically include descriptions of general species tion, have escaped expected natural disturbances and are composition, changes in community units and reference to atypical of the ÔcommonÕ landscape at any time. The few blazed trees marking predetermined points at intervals along uncut stands are a limited and selective spatial sample and the survey lines (e.g. Bourdo, 1956; White, 1984; Manies & survived exactly because they have unusual histories or Mladenoff, 2000). In contrast, in colonial New England and extreme settings (Cogbill, 1996). Predictable physiological later in the original states, town-mediated surveys regularly and silvicultural responses are dependent on temporal conti- cited only ÔwitnessÕ trees. These unregulated and unstan- nuity and stable environmental conditions, which are con- dardized eastern surveys have received relatively little inter- strained by species migrations, climate changes and soil est, perhaps because they are obscure and are found in widely development in glaciated regions (Russell, 1997). Models tend scattered repositories (e.g. McIntosh, 1962; Russell, 1981; to be simplistic, deterministic and linear expressions of a few Loeb, 1987; Seischab, 1990; Marks & Gardescu, 1992; common stereotypes. Thus the current vegetation in New Abrams & Ruffner, 1995; Cowell, 1995; Abrams & McCay, England is potentially a biased evidence of the past and the 1996; Cogbill, 2000; Black & Abrams, 2001). In New Ó 2002 Blackwell Science Ltd, Journal of Biogeography, 29, 1279–1304 Presettlement New England 1281 England, case studies have analysed the presettlement com- tially transgressive, regularly progressing from the southern