Assessing Native American influences upon pre-European tree species composition in western New York: An approach using original land survey records and species distribution models
Tulowiecki, Stephen James
MetadataShow full item record
The impact of Native Americans upon forest composition has long been a topic of scholarly inquiry and debate. Native American groups of Eastern North America practiced various forms of horticulture and land management (e.g. forest clearance and burning), which had the potential to alter forest composition. However, questions surround the degree to which Native Americans altered forest composition, and the spatial extent of these alterations. To understand Native American impacts, researchers have utilized presettlement land survey records (PLSRs), which are the records of the first land surveys by European-Americans in North America. PLSRs are invaluable to researchers, because PLSRs contain descriptions of forest composition along survey lines ("line-description" data), as well as the records of trees that surveyors blazed adjacent to survey monuments ("bearing-tree" data). Researchers have applied the vegetation data within PLSRs to develop species distribution models (SDMs), in order to model the past geographic distributions of tree species. However, no research has utilized PLSRs and SDMs together, to understand Native American impacts upon past forest composition. This dissertation contains two purposes. The main purpose is to utilize PLSRs and SDMs, in order to assess the degree to which Native American land-use practices altered forest composition, in the era just prior to European-American settlement ("presettlement"). A second purpose of this dissertation is to explore data quality and modeling issues surrounding the applications of PLSRs for developing SDMs. Using PLSR data from western New York, USA, this dissertation addresses its purposes in three studies. Chapters Two and Three explore issues surrounding use of PLSR data for developing SDMs. Using Holland Land Company (HLC) data (circa 1797-1799 CE) in western New York, Chapter Two first quantifies the positional error that is associated with different methods for georeferencing PLSR data, and then examines how this positional error impacts the performance of SDMs that model the past distributions of tree species. Chapter Three explores which PLSR datasets produce SDMs of the highest predictive performance, using the bearing-tree data and line-description datasets of the HLC lot surveys (circa 1799-1814 CE) as training data. Chapter Three also explores whether HLC surveyors were biased in their selection of bearing-trees. Chapter Four addresses the main purpose of the dissertation, by understanding the impacts of Native American land-use practices upon presettlement forest composition in Chautauqua County, New York. This study uses PLSR data and SDMs to model the presettlement distributions of tree species, as a function of environmental and Native American influences. The study utilizes historical and archaeological data to develop "Native American variables" (NAVs) as predictors of species distributions. Chapter Four assesses (1) whether NAVs improve the predictive performance of SDMs; (2) how important NAVs are to SDMs; and (3) the spatial extent of Native American impacts upon species distributions. Chapters Two and Three yield four notable results, regarding the joint use of PLSRs and SDMs. First, Chapter Two manifests that positional error in PLSR data can negatively impact the predictive performance of SDMs, and can also lead to the selection of different predictor variables of species distributions. Second, Chapter Two suggests that the impact of high positional error upon the predictive performance of SDMs is most severe for species typically found in marginal environmental conditions, relative to the study area. Third, Chapter Three suggests that line-description data produce SDMs that more accurately predict species presence and absence in independent datasets, in comparison to SDMs trained from bearing-tree data. Fourth, Chapter Three manifests that surveyors were biased in their selection bearing-trees, on the basis of both taxon and diameter. Chapter Four produces three important results, regarding Native American impacts upon tree species distributions. First, knowledge of Native American settlement improves our understanding of tree species distributions prior to European-American settlement in Eastern North America. Second, the results show that at the local- to regional-scale, NAVs are some of the most important predictors of presettlement tree species distributions, especially for mast-bearing (i.e. producing large nuts) tree species. Third, results suggest that Native American settlement increased the spatial distribution of mast-bearing tree species. The results of this dissertation collectively demonstrate how PLSRs and SDMs can be synthesized to understand the role of Native Americans in shaping the past distributions of tree species in Eastern North America. Beyond the immediate findings, the results of this dissertation provide an ecological baseline against which to measure future landscape change within the study area. This study also provides a methodological template for understanding the distribution of culturally-modified landscapes prior to European-American settlement, and is informative in shaping modern conservation and restoration goals that seek to reflect prior ecological conditions.