WITH TREES AS MY WITNESS: The Use of Historical Survey Records in Understanding Contact-Era Forest Structure in The Upper Middle Fork Willamette River Watershed by GRACE PETRAGLIA A THESIS Presented to the Department of Environmental Studies and the Robert D. Clark Honors College in partial fulfillment of the requirements for the degree of Bachelor of Science May 2023 An Abstract of the Thesis of Grace Petraglia for the degree of Bachelor of Science in the Department of Environmental Studies to be taken June 2023 Title: With Trees As My Witness: The Use of Historical Survey Records in Understanding Contact-Era Forest Structure in The Upper Middle Fork Willamette River Watershed Approved: Michael R Coughlan, Ph.D. Primary Thesis Advisor This research examined the use and limitations of United States General Land Office surveys for reconstructing historical landscape patterns. We reconstructed forest structure for three townships surveyed in 1869 and 1902 in the Upper Middle Fork Willamette watershed. Survey of the area was preceded by early Euro-American explorers, construction of the Oregon Central Military Wagon Road, and timber and grazing activity. We used historical survey notes to conduct an iterative fuzzy classification of section lines that assigned each line degree of membership to closed canopy forest. The final fuzzy classification relied on a composite index comprised of the number of witness and bearing trees recorded along each line and the average distance of each tree recorded from the line. This approach minimized research bias and mitigated the inherent limitations of historical data. Fuzzy membership was then compared with other variables derived from survey notes as well as a previously published reconstruction that uses discrete vegetation classes. Our findings show a positive relationship between survey line membership to degree of closed canopy forest and elevation. We conclude that fuzzy set theory is an appropriate method to work with GLO survey records. However, our investigation suggests that on their own, GLO survey data is insufficient in quality and reliability to confidently reconstruct historical forest structure at a spatial grain needed to inform management plans. 2 Acknowledgements My biggest thanks to my primary advisor Dr. Michael Coughlan for giving so selflessly his time and energy to this thesis. Mike, thank you for your infinite patience and guidance throughout this entire process. Thanks to my CHC advisor Dr. Barbara Mossberg for her kind words of encouragement and passion for this topic. Many thanks to my family for their belief in me and to my friends and partner for their constant support! 3 Table of Contents Introduction 7 Public Land Survey System 8 Background 10 History of GLO Survey Records in Research 12 Reckoning with Biases and Limitations 14 “Pre-settlement” an Accurate Description? 15 Site Specificity 16 Current Use of GLO Surveys 17 Methods 19 Study Site 19 Survey Note Transcription 19 Data Exploration 21 Analyzing Forest Structure 22 Fuzzy Set Construction 24 Mapping the Fuzzy Sets 26 Results 27 Discussion 36 Comparing Christy and Alverson’s Willamette Valley Historical Vegetation Map 36 Implications for Future Use of GLO Survey Records 38 Recommendations for Further Analysis 39 Conclusion 41 Bibliography 56 Supporting Materials MS Excel: GLO_Coding_ThesisCopy 4 List of Figures Figure 1: Project area by township and range 20 Figure 2: Schematic of a section line 21 Figure 3: Fuzzy “Or” Method 28 Figure 4: Fuzzy “And” Method 28 Figure 5: Fuzzy “Sum” Method 29 Figure 6: Comparison of bearing trees of resurveyed lines 34 Figure 7: Comparison of DBH of resurveyed lines 34 Figure 8: Comparison of number of different tree species on resurveyed lines 35 Figure 9: Discrete versus fuzzy classification 38 5 List of Tables Table 1: “Heavy timber” example survey lines 23 Table 2: Excluded variables 31 6 Introduction The forests in the western United States are challenged by a slew of factors that limit their resiliency: history of fire suppression, timber harvest, pestilence, and droughts and wildfires linked to climate change (Levine, et. al., 2017). In response to these stressors, forest managers are turning to resilience-based management plans that model conditions before Euro-American colonization. The forests before Euro-American colonization are thought to have been healthier and more resilient than contemporary forests and as such, should form the basis of current and future management plans. Research into General Land Surveys (GLOs) has been increasing in the past several decades to reconstruct those historical forests. The limitations and biases inherent in the survey data are widely acknowledged by researchers, but those factors do not act as a hindrance. The objective of this research is to evaluate the scientific reliability and validity of GLO data as a source to understand historical forest conditions. To address our objective, we applied fuzzy set classification to historical land survey notes dating from the period of Euro- American colonization (Contact Era) to reconstruct forest structure in a portion of the upper Middle Fork Willamette River watershed. We devoted significant time to data exploration and analysis to ensure minimal researcher subjectivity and maximum confidence in our methods. We used section lines as the unit of analysis and kept our classification general with degree of closed canopy forest, instead of extrapolating discrete vegetation type boundaries. The analysis conducted in this study consists only of data directly from the GLO surveys, we did not supplement the classification with other information. Our conclusions show that as a baseline, GLO surveys provide valuable insight; however, the reliability depends on the methodology and spatial grain of the investigation. 7 Public Land Survey System In 1785 the U.S. government established the Public Land Survey System (PLSS) to map and divide land for sale and settlement across the entire United States (Whitney and Decant, 2001). From Ohio to the West Coast, a rectangular system, colloquially known as the township and range system, was used to conduct the land surveys. The General Land Office (GLO) was the lead department and contracted thousands of surveyors to carry out the work. Survey of Oregon began in 1851 with the establishment of the Willamette Meridian and the baseline; at the intersection of these lines is the initial point from which all surveys began (Stewart 1935). While the purpose of the PLSS was to provide a systematic framework for surveying and dividing land as real property; the true work of surveyors was to assess the land quality and potential uses, and physically mark boundaries between sections linking the legal representation of land embodied in the paper map to the land (i.e., real property). The basis of their notation and marking system was the assignment of specific trees as “witness trees” for locating and re- locating section boundaries. As surveyors traveled along compass bearings, they selected and recorded trees to mark (or “witness”) the midpoint and beginning/end of each section line (i.e. section corners). At the midpoint, surveyors chose two trees, one for each section, to mark with the township, range, and section number. Section corners serve as both beginning and ends of lines, so surveyors preferably selected four trees, one for each section, and marked them accordingly. For each tree, surveyors noted the species, diameter in breast height (DBH), and direction and distance from the midpoint or corner. Surveyors were also required to record information like cultural features (roads, trails, burns), terrain, soil quality, and understory flora. Soil quality was rated on a scale 1-4 with 1 being high quality and 4 being poor quality. 8 This information was collected to assess the potential use and value of the land to Euro- American society; section lines that passed through low, flat, and mid to low grade soil were noted for potential settlement, while section lines rated for first rate soil were noted for potential agricultural use (White, 1926). 9 Background The spatial extent of the study for our investigation is along the Middle Fork of the Willamette River in the Western Cascades at a point just upstream from Oakridge, Oregon. The study area is in the foothills of the southern end of the Willamette Valley: a rich and fertile region bounded by the Cascade, Coast, and Calapooya ranges. Because the townships are mountainous and were not considered an ideal place for settlement or other settler colonist activities of commerce or production, the area was not surveyed until almost two decades after survey of Oregon began. The wealth of resources and diversity of life in the valley supported indigenous peoples for thousands of years before Euro-American arrival and colonization. A common misconception, that has by now hopefully been dispelled by research produced on the subject, was of the landscape as being pristine before Euro-American colonization. In truth, extensive scholarship has brought to light the dynamic relationship that existed between indigenous peoples of North America and their environment–a relationship that shaped the landscape that settler colonists first saw (Boag, 1992; Boyd, 1999). Indigenous fire use maintained the open prairies and savannas of the valleys which benefited the tribes by drawing game animals and increasing production of food crops (Kimmerer and Lake, 2001). It is these kinds of disturbance regimes and vegetation patterns that researchers seek to understand with GLO surveys. Presence of European fur trappers and explorers had been documented in the valley as early as the 1700s, and they brought disease with them that decimated indigenous populations before major westward expansion in the 1840s (Olson and Suski, 2021). Timber production has been an important sector in the Pacific Northwest since the arrival of European settler colonists to the region; therefore, the study of forest management plans and 10 policy is a justified and relevant topic for understanding the changes to the landscape. Management changes since the incorporation of what is now called the Willamette National Forest into the national forest system has shaped the land use history of our study area. One of the first models of forest management was sustained yield which was developed under Pinchot conservation: prioritization of use over preservation (Johnson and Swanson, 2009). Sustained yield was characterized by clear cutting, fire and pest suppression, and late successional harvesting to make way for young faster growing stands. Clear cutting was the prescribed method for the Middle Fork Willamette watershed which contributed to increased landslides in the area (Lyons and Beschta, 1983). This model valued forests and timber as a commodity and did not acknowledge their ecological value. An estimated 300 million board feet were produced in the Willamette National Forest alone from 1930 to the 1950s (Tonsfeldt, 1994). Timber harvest was concentrated in the Cascade foothills at sites like our study area where easy access was facilitated by roads like the Oregon Central Military Wagon Road. A significant shift in management came during the 1970s with the passage of the National Environmental Policy Act (1970), Endangered Species Act (1973), and National Forest Management Act (1976). Collectively, these pieces of legislature changed the approach from conservation to ecological forestry. One of the most significant moments in forest management history was the Northwest Forest Plan (NWFP) of 1994 spurred by the controversy over the logging of old growth forest, primary habitat for the endangered northern spotted owl (Strix occidentalis caurina). The NWFP plan area was bounded by the range of the owl in the federal forest lands of Northern California, Oregon, and Washington. This plan was more ambitious in its goal to preserve forest lands and increase health of the various forest ecosystems to maintain diversity of native populations. 11 Within the Middle Fork NWFP project site, the plan prescribed thinning services, some for timber harvest and some not, to decrease fuel abundance and therefore reduce fire severity (Thomas, et. al., 2006). Restoration in the Middle Fork area continued use of thinning, bringing back meadow habitat, and repairing the riparian corridor (Franklin and Johnson, 2012). In the last two decades since its passage, the focus of the NWFP has shifted from conservation and restoration to the revival of historical forest disturbance regimes, namely low intensity fires (Spies et. al., 2019). Not long after, the Healthy Forest Restoration Act was passed in 2003, mostly in response to recent catastrophic wildfires and the growing realization of the role forest management played in them. According to Johnson and Swanson (2009), the act set two important precedents: i. The first statutory protection of old-growth forests, and ii. The first statutory call for the use of past ecological conditions in planning. (p. 15) The second precedent keyed forest managers into the need to understand historical forest conditions; thus, the increased interest to use GLO records to inform management plans. Some concern over this law was that fuel reduction was a guise for timber companies to increase their access (Hanson, 2002); however, the evidence to substantiate the claim is unclear. Thoughtful and intentional management of forests is of increasing importance as wildfires become more severe, rural communities are put at risk, livelihoods are threatened, and natural habitat is shrinking. Researchers continue to develop ways in which data from GLOs can best be utilized to provide solutions. History of GLO Survey Records in Research Use of GLO surveys in research has a long history but wasn’t widely employed in historical landscape reconstruction studies until the mid-20th century. The first documented use 12 of GLO surveys in research was by naturalist Increase Lapham in 1855; Lapham used the survey notes to reconstruct locations of windfalls, prairies, and other features in Wisconsin to understand the effect of disturbance regimes on vegetation (Dorney, 1983). It was nearly three quarters of a century later in 1925 when ecologist Paul Sears used GLOs to define historical prairie boundaries and distribution of tree species across the state of Ohio (Sears, 1921; Sears, 1925). At the time of Sears’ investigation, the survey of Ohio was not yet complete; therefore, as many future studies would do, he supplemented the land surveys with geological surveys and qualitative sources such as early explorers’ journal entries. The methodologies Sears applied in his research were influential for many other researchers, namely those on the northeast coast (Stuckey, 2009). Use of GLOs was still slow to gain traction with ecologists and others in the academic circles until Bourdo’s paper A Review of the General Land Office Survey and of Its Use in Quantitative Studies of Former Forests (1956) which encouraged the use of land surveys in ecological reconstruction work (Dorney, 1983). Prior to the establishment of the rectangular system, land surveys on the east coast began as “colonial surveys” that recorded only acreage and vegetation type (i.e., woodland or meadow), and later transitioned to the metes and bounds system (Whitney and DeCant, 2001). Previous work also depended on qualitative sources like journal entries from early explorers which were often exaggerated and generally less reliable (Noss, 1985). As a semi-standardized source with both qualitative and quantitative data, ecologists and other researchers quickly took to incorporating GLO surveys into their work. A Google Scholar search (accessed May 15, 2023) shows that between 1920 and 1956 (the year Bourdo published his paper), there were just 75 records of “pre-settlement vegetation” papers; from 1956 and until 2000, the number of papers retrieved by the Google search was almost 7,000, and for the last two decades that number is 13 more than double. With this increasing interest in reconstructing historical landscape conditions, it is of increasing importance to thoroughly understand and evaluate GLO surveys strengths and weaknesses as a reliable data source. Reckoning with Biases and Limitations Despite the widespread use of GLO surveys, many of the limitations and biases of the data are broadly acknowledged. Much of the uncertainty around GLO surveys as a dependable source stem from inconsistency of survey handbook rules and a few known cases of surveyor fraud. Survey rules were still being defined in the first several decades following the creation of the PLSS which created variation in survey documentation; for example, standards for selecting trees at quarter sections changed from trees at “adjacent sections” to merely “different sections” between 1831 and 1850 (Bourdo, 1956). The vague language and back and forth changes likely created confusion and jeopardized consistency within surveys. In addition, the level of detailed notation required by the surveyors changed throughout the decades. For instance, the enumeration of plant species was specified in 1843, and natural disturbances and water feature boundaries were required in 1850 (Whitney and DeCant, 2001). The rules were finally standardized in 1855 (White, 1984). Furthermore, the effect of instances of fraud on the data, although speculated to be uncommon and remedied by the surveyor generals, cannot be ruled out entirely. Glaring examples of fraud in which surveyors fictionalized accounts of survey lines were usually quickly discovered and the lines resurveyed; one such resurvey revealed missing corners and erroneous distances between the survey line and landmarks (Stewart, 1935). Nonetheless, researchers often acknowledge the difficult conditions of the surveyors’ work and commend the fact that most of the surveys were well done (Schulte and Mladenoff, 2001). 14 The uncertainty surrounding surveyor reliability and survey instructions bleeds into all areas of the data. One central concern is the ability of land surveys to support quantitative versus qualitative analysis. Researchers express caution against the use of land surveys for quantitative analysis because of low confidence in accurate survey methodology (Grimm,1984; Wang, 2005). This is also doubt over the extent to which any findings could be compared to modern forest inventories. Formulas have been created to account for surveyor bias in quadrant placement and bearing angle (Kronenfeld and Wang, 2007; Hanberry et. al., 2012), and tree density estimates (Pollard, 1971; Jost, 1993). Numerous methodologies have likewise been developed to extrapolate point data of bearing trees into continuous vegetation coverage (Brown, 1998; Manies and Mladenoff, 2000; Fagin and Hoagland, 2011). The many limitations and uncertainties regarding GLO survey data are well documented and addressed but do not impede their continued use. “Pre-settlement” an Accurate Description? As one of the few systematic historical records of landscape conditions, GLO surveys are most commonly used to understand vegetation patterns before Euro-American colonization. Without justification or reference, “pre-settlement” is the term most often used to describe that time period (Galatowitsch, 1990; Kronenfeld and Wang, 2007; Aube, 2008); however, this term may not be the most accurate description of historical landscape conditions. In many cases, there were pre-existing Euro-American settlements or extensive colonial commercial activities in areas that had yet to be surveyed (Horne, 2010). For example, commercial logging in Maine began in the 1650s, but the earliest survey records were not completed until 1793 (Lorimer, 1977). In addition, Freidman and Reich (2005) described 1880s Minnesota as pre-settlement even though population of Euro-American settlers in area boomed in the mid-1800s (Larsen, 1940). The term 15 “pre-settlement” references the period prior to Euro-American colonization; therefore, to use this term would be inaccurate in many cases. The GLO survey notes are often the oldest systematic records of landscape conditions and therefore the closest proxy data for pre-settlement vegetation conditions. However, the term is rooted in the Eurocentric idea that prior to Euro-American arrival and colonialist activities, North American landscapes were “pristine,” thereby neglecting any recognition that landscapes were shaped by thousands of years of indigenous inhabitation and stewardship (Williams, 2000; Kimmerer and Lake, 2001). Euro-American presence in Oregon was limited to fur trappers and a few early settlers until an influx of people arrived via the Oregon Trail in 1843 (Boag, 1992). Settlers concentrated in the fertile lands of the northern Willamette Valley before spreading outward and down. Construction of the Oregon Central Military Road was one of the drivers for workers and their families to move south towards the mid-1850s (Beckham, 1981). Although the foothills of the Cascades were not conducive to settlement, there was substantial colonialist activity taking place. In our study area, research shows that commercial grazing was introduced around the time of survey in 1869 which likely impacted the environment observed by the surveyors (Hadley, 1999). Furthermore, settlers’ desire to forge a trail south for settlement of the Klamath Basin had explorers scouting routes in the Western Cascades in the 1850s before land surveys in Oregon began (Beckham, 1981). Therefore, Contact Era is a more accurate description of the temporal setting of our research rather than “pre-settlement” which gives a false sense of historical human- landscape relationships. Site Specificity Considering the variable nature of surveyor instructions, known cases of fraud, and inconsistent settlement patterns, researchers must conduct thorough investigation into their study 16 area. Indeed, independent studies have been dedicated to understanding the instructions of one’s state or region (Dodds, et. al., 1943). Depth of knowledge of any instruction inconsistencies or individual surveyor biases will help inform appropriate methodologies based on the unique limitations of one's study area (Manies, et. al., 2001). Areas surveyed using the metes and bounds system, which featured unevenly distributed survey sections, would necessitate a different approach (Aube, 2008). Methodologies to assess bias are likewise impacted by survey differences. Bourdo (1956) analyzed distances between bearing trees and wooden stakes used to mark section corners to evaluate tree selection bias; however, no such markers were used in Maine surveys, so Bourdo’s method cannot be applied to surveys from Maine (Lorimer, 1977). Much of the existing research takes into consideration how the unique characteristics of the study site impact data limitations. Current Use of GLO Surveys Over the decades, researchers have explored various methods and approaches to GLO data interpretation. GLO records have been used for large-scale vegetation reconstructions (Christy and Alverson, 2011; Powell, 2008), forest composition analyses (Friedman and Reich, 2005; Knight, et. al., 2020), and comparisons to modern day forests (Dryer, 2001; Hessburg, et. al., 2005). Some methodologies took a deductive approach to create a coding schema and pulled information from supplementary sources such as plat maps and soil surveys (Habeck, 1961; Christy and Alverson, 2011). While this approach maintained consistency within classifications, pre-defining classes may have obscured patterns that otherwise might have been seen. Other methodologies conducted analyses with a more complex statistical approach, for example estimation of forest density (Anderson, et. al., 2006; Hanberry, et. al., 2012; Levine, et. al., 2017), and fuzzy sets (Brown, 1998). Brown critiqued the assignment of discrete boundaries on 17 vegetation types because that assumes that only one vegetation class can exist at a given location. Instead, Brown used fuzzy membership, which is the transformation of data along a common scale (usually 0 to 1) based on degree of membership to a set. This method allows a many-to- many relationship between section lines and degree of membership to closed canopy forest. Regardless of the approach taken, studies of GLO survey records continue to be published and discussed because, in most cases, they represent the only systematic record of historic forest conditions that exist (Schulte and Mladenoff, 2001). Overall, scholars come to a shared conclusion that imperfection is not reason enough to reject a valuable data source on historic land cover. GLO survey data was collected for the specific reason to assess the potential value of land for settlement, farming, and timber harvest, so data must be carefully interpreted when used for other reasons. Based on settler activity in our study area, the difficulty of traversing mountainous terrain, and the other limitations we have listed above, we chose a methodology that embodies all the uncertainty of working with GLO data, but that is still founded on confident variable relationships: fuzzy set classification (Brown, 1998). Fuzzy set classification is an appropriate research approach to inform the reconstruction of historical forest structure from survey records because it accounts for some of the inherent limitations of the data. 18 Methods Study Site The project area consists of the upper Middle Fork Willamette River watershed located within Township 23 South Range 3 East, and Township 24 South Ranges 3 & 4 East (Figure 1). These sections were chosen because 1) the Forest Service is actively conducting restoration projects in the area, and 2) the Oregon Central Military Wagon Road runs through the area which resulted in a relatively early government survey by the General Land Office. The earliest surveys for the area were conducted in 1869 and 1902; thus, we selected these surveys to capture the landscape conditions as close to the Contact Era time frame as possible. Contact Era is defined as the time of first arrival by European settler colonists to land occupied by indigenous peoples; the Oregon trail facilitated the influx of colonists’ arrival to Oregon in the 1840s. The arrival of European colonists through the Oregon trail accelerated changes to the landscape that were already set in motion by the forced removal of indigenous groups from the valley just a few decades prior. Survey Note Transcription Scans of original surveyor notes were downloaded from the Bureau of Land Management (BLM)1 and organized by township and range, year, and surveyor. Survey pages were cataloged by section boundaries (North, South, East, West, or subdivision) which were then combined by township into one document (pdf file format) so that the surveys could be transcribed chronologically. We transcribed survey notes from the PDF files into an MS Excel spreadsheet so that data could be pulled out and analyzed. Although the 1902 notes had been transcribed and 1 https://www.blm.gov/or/landrecords/survey/ySrvy1.php accessed and downloaded 6 July 2022 19 typed some time prior to digital archiving, the notes from the 1869 surveys were scanned and archived by the BLM in their original form of handwritten script; therefore, data transcribed from these scans into MS Excel included a column for confidence level to account for the difficulty of reading handwritten notes. We took an iterative approach to transcription; as the handwriting of each surveyor became more familiar, we were able to return to areas of uncertainty and make updates and corrections. Figure 1: Project area by township and range Note. Project area is shown in yellow color overlaid by the township and range grid. Townships are divided into 36 sections of one-by-one mile squares. Surveyors completed a survey of a square section going line by line (e.g., 4 lines for each side of the square, unless one of the lines had already been surveyed for the previous section). Surveyors used a measurement system with units called chains and links to mark their course: a chain had 100 links and totaled 66 feet in length (Powell, 2008). A section line is 80 chains long (i.e., one mile). Figure 2 shows what a section line crossing the Middle Fork of the Willamette River may have looked like based on survey data. 20 For data entry, we set up a spreadsheet so that each row corresponded with a point on the section line (e.g., 40 chains from the start corner) at which a surveyor noted a tree, change of topography, or other point of interest. The surveyor also noted the distance to the feature from the line (e.g., 10 links), feature characteristics such as tree species, stream width, or rock type, and if it was a tree, the diameter at breast height (DBH). For each row, we also documented the township and range, the year, the direction the surveyor traveled on the line (e.g., East to West), and the name of the surveyor. We assigned a unique ID to each section line to more easily track which features belonged to which section lines. When surveyors reached the end of a section line, they often described the overall characteristics of the land: terrain, soil quality, stand structure, tree species, and undergrowth. In anticipation of later analysis, we calculated the average and maximum DBH and distance of bearing trees for one section line. To capture this information pertaining to the entire section line, we created a separate spreadsheet with the individual rows corresponding to entire section lines. We linked this data with the individual features along a line through the unique ID. Figure 2: Schematic of a section line Data Exploration The variable types transcribed from survey records include ordinal (soil rating), categorical (tree species), continuous (measure of distance and DBH), and counts (number of bearing trees). Variable relationships were explored using histograms, bar graphs, box plots, and 21 regression analyses. Terrain, soil description, and stand structure were ranked and assigned a value; in some cases, similar descriptors such as “stony” and “rocky” were combined into one category. If multiple soil ratings were recorded for a section line, we took the average. Understory and tree species were converted into a binary variable to represent presence or absence of specific species (or species families as specified in the notes). Because the surveys used in this project come from both 1869 and 1902 resulting in a 33 year gap between surveys, we decided to separate the data and explore the variables as two sets. During this process, we identified lines that were originally surveyed in 1869 and were resurveyed in 1902 for comparative analysis. Some standalone section corners that had been resurveyed were likewise identified. To assess differences in survey methodologies and forest changes between 1869 and 1902, we compared the number of witness trees, the DBHs, and tree species of the original and resurveyed lines. Resurveyed lines were excluded from the 1902 portion of the data set for forest structure analysis and maps in favor of representing the earlier, 1869 data. Analyzing Forest Structure Initially, at the start of this project, we considered adopting Christy and Alverson’s 2011 coding schema, but we found it to be a poor fit for the data from either of the surveys. Their methodology used distance to bearing tree and stand structure and undergrowth descriptors to categorize survey lines into crisp categories of forest, woodland, savanna, or prairie. For example, their schema dictated that a line be classified as “forest” if all bearing trees were present and within 100 links of the line, and if terms like “heavy”, “dense”, or “brushy” were used. We found it difficult to match lines to those ecosystem categories based on this coding schema. For instance, the range of BT distances we transcribed was 1.36 to 258 links, but only 2% of bearing tree distances had a value over 100 links; therefore, distance as a classification 22 tool made little sense. A review of statistical comparisons to understand variation in surveyor performance suggests a preference for trees located closer to the line for reasons of time, pay, and the difficult nature of the work (Schulte and Mladenoff, 2001). This cast further doubt on the meaning and reliability of the distance metric as used in Christy and Alverson’s coding schema. Furthermore, of the 136 lines that had stand structure descriptions, “heavy timber” made up 90% (122 lines). Under the Christy and Alverson schema, a section line with three bearing trees all under 100 links from the line and a descriptor of “heavy timber” fits all but the bearing tree requirement to be coded as “forest”; however, we felt that a line with this description is more likely to have had a more open structure than a closed one, rendering the “forest” designation misleading (Table 1). In addition, as noted below, Christy and Alverson apparently did not follow this scheme very closely in their reconstruction of our project area, but for undisclosed reasons privileged historical aerial photos. Table 1: “Heavy timber” example survey lines Section Township and Survey Stand Tree Understory # Distance Comment Line Range Line Structure Species Species BTs (links) ID 375 T23S_R03E_1902 E-W between heavy fir, vine maple, 6 55, 46, 33, NA S20-S29 timber hemlock rhododendron 28, 13, 10 139 T24S_R04E_1869 W-E between heavy not not noted 3 14, 27, 30 NA S1-S12 timber noted 189 T23S_R04E_1902 S-N between heavy pine, salal & 10 20, 89, 30, mentions T23SR3ES25 timber larch & huckleberry 13, 39, 76, open prairie - fir 20, 24, 6, at 30 links T23SR4ES30 16 Note. None of these lines fit into the Christy and Alverson coding schema. We found their approach to be too rigid to explain something as dynamic and complex as forest structure. 23 Fuzzy Set Construction A fuzzy set is a group (set) of objects that have varying degrees of membership to the set (Zadeh, 1965). Fuzzy set theory is useful when working with data that is imprecise and can accommodate both quantitative and qualitative data. We normalized data by scaling variables between 0-1 where 0 is full non-membership and 1 is full membership, and any value in between describes the degree to which an object is a member of that set. We used min-max normalization where the minimum and maximum were taken from the overall dataset; the min and max were kept the same to normalize data from both 1869 and 1902 survey years. We used the formula below shown in (1). Fuzzy membership is not to be misunderstood as simply the transformation of variables along a continuum, but a process based on thorough understanding of relationships and a sound knowledge base (Ragin, 2000). Our objective was to use the GLO survey data to construct a fuzzy set for closed canopy forest structure to help inform better understanding of Contact Era forest structure. 𝑥𝑥 − min (𝑥𝑥) 𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋𝑋 = ( ) (1) max 𝑥𝑥 − min (𝑥𝑥) We identified the following variables for possible inclusion in the fuzzy set based on careful consideration of how they relate to forest canopy closure: the number of bearing trees per section line, the median, average, and maximum inverse distance of the bearing trees from the line, and the tree species and undergrowth communities. The number of bearing trees was a clear choice; we trust that there were no (or few living and of sufficiently robust size) trees present when the surveyor notes indicate so, and vice versa. Although we had difficulty using distance in parameters used by Christy and Alverson (2011), we felt that distance of the bearing trees from the line does contribute to our overall understanding of forest structure. We suggest that shorter 24 distances between the line and the bearing tree would indicate a more closed canopy forest; however, since larger values on the 0-1 scale indicate higher degrees of closed canopy, we took the inverse value of distance (1-x) as a pre-processing step before normalization. Therefore, we chose to include the number of bearing trees and average inverse distance of the bearing trees from the line for both 1869 and 1902 as indicators of closed canopy forest. The minimum and maximum values found were 0 and 12 for number of BTs, and 0 and 258 for distance. The variables were normalized separately, then combined in three different ways to create three versions of the fuzzy index of closed canopy forest: 1) the fuzzy “and,” the minimum of the input, 2) the fuzzy “or,” the maximum of the input, and 3) the fuzzy “sum,” the average of input (ArcGIS Pro). (See Appendix I) We also analyzed tree and understory species for 1902 as indicators for degree of forest canopy closure. The Field Guide to the Forested Plant Associations of the Westside Central Cascades of Northwest Oregon from the United States Forest Service was used to attribute a percentage of canopy closure to each pairing of tree species and undergrowth (McCain, 2002). Since the surveyors only recorded tree species as “fir”, “pine”, “hemlock” and not the specific species, the grouping of tree and understory species for each line was matched as best as possible to the plant associations in the field guide. The guidebook mapped plots where each association was found; therefore, we gave priority to associations which had plots nearest to our project site. Additionally, we gave preference to associations with the highest presence percentage of the understory species. For lines that had several possible associations, we took the average percentage of canopy closure. Because of the high uncertainty, we ultimately excluded tree and understory species from the final fuzzy set (see Appendix II). 25 We gave priority to the oldest available survey; therefore, a section corner with no BTs recorded in 1869 was maintained as having zero BTs, even if a 1902 resurvey recorded four BTs. Surveys from 1869 had less data recorded; understory species were “not noted” for all but a few lines for which surveyors noted “some grass”, and there were no soil descriptions recorded. Some lines had no tree species recorded in their line summaries but had BTs; in this circumstance the BT species were substituted for their line summary. We felt that BT species were a reasonable proxy for these lines; however, it is important to note that tree species listed and the species of the BTs of the same line do not always correspond in the original survey notes. Furthermore, during preliminary iterations of visualizing the 1869 fuzzy set, we noted that lines along the Middle Fork of the Willamette River were coded as being more closed than open canopy. In addition, a survey line following the course of the river indicated that the riparian corridor was apparently characterized by a closed canopy forest. These data points skew the fuzzy value towards canopy closure for each line crossing the river. To account for this, we removed the BTs that were recorded within the riparian corridor from the fuzzy analysis. Mapping the Fuzzy Sets We mapped the fuzzy sets along the GLO survey lines in ArcGIS Pro 3.0.2. Degree of membership to closed canopy forest was symbolized using the unclassed method, and the 1869 data is distinguished from 1902 by solid and dotted lines, respectively. Unclassed colors symbology does not use discrete classes like graduated colors but rather evenly assigns colors across the input value range (ArcGIS Pro); we felt unclassed colors best represented the fuzzy index for closed canopy forest. 26 Results The three fuzzy sets of closed canopy forest are shown below (Figures 3-5). The spatial distribution of section lines categorized by degree of membership to the closed canopy forest shows a gradient between less closed canopy lines and more closed canopy lines. Across the three methods— “or”, “and”, and “sum” (see Appendix I)—lower degrees of closed canopy forest are characterized by the lighter shades of orange as seen along the Middle Fork of the Willamette River course. The river course is situated at lower elevations and the topography is relatively level and more gently sloping. In comparison, the survey lines modeled as higher degree of membership to closed canopy (in the darker shades) are located at higher elevations in the mountainous terrain surrounding the river corridor, most of which were surveyed in 1902. Most lines produced using the “or” method show more closed canopy forest outside of the riparian area (Figure 3); “or” maximized the input fuzzy membership which tended to favor the fuzzy variable for distance of bearing trees from the line. Because shorter distances suggest a more closed canopy and the inverse distance was taken to transform the value along the 0-1 index, the fuzzy variable for distance produced higher index values. The map produced using the “and” method displays the most variability of degree of closed canopy (Figure 4), but still follows the overall pattern of less closed canopy forest along lower elevations and more closed canopy in the uplands, generally >3,600 meters in elevation. The “sum” method maintains some of that variation but overall shows higher degrees of closed canopy than the “and'' method because it took the average of the fuzzy variables for distance and number of bearing trees (Figure 5). 27 Figure 3: Fuzzy “Or” Method Figure 4: Fuzzy “And” Method 28 Figure 5: Fuzzy “Sum” Method Excluded variables from the final indices for degree of closed canopy forest are shown in Table 2. Poisson regression of average DBH and number of bearing trees had a p value < 0.05 and a pseudo-R squared value of 0.25; DBH may have been included if multiple studies hadn’t already exposed tree selection bias (Bourdo, 1956; Noss, 1985; Kronenfield and Wang, 2007). We selected fuzzy “sum” to compare and explain the exclusion of the other GLO variables; the other two methods favor the lower or higher values, but “sum” takes a combination which we felt presented a more holistic view. The box plot of terrain and the fuzzy index shows that a wide range of degrees of canopy closure correspond to lines described as “level”, “slopes”, “rolling”, and “broken”; all first quartiles are 0 and the third quartiles range from 0.52 to 0.69. Comparatively, lines which were described as “mountainous” have a much smaller interquartile range with an overall higher degree of closed canopy: first quartile 0.70, third quartile 0.86, and median value of 0.83. The differences between the box plots for “level” and “mountainous” roughly correspond to the distribution of less and more closed canopy across the topography of 29 the project area; along the riparian corridor the land is flatter, and the forest canopy is less closed, and the surrounding area is more mountainous with a more closed canopy forest. A box plot of stand structure shows that “heavy timber” does apparently coincide with the fuzzy index; the interquartile range is 0.70 to 0.86, the median value 0.81, and the mean 0.76. The box plot for “open” has a larger interquartile range– 0 to 0.61–but still has low median and mean values: 0 and 0.24 respectively. If we look at the lines described as “open”, the indices do coincide with that description: line IDs 111, 113, and 153, for example. Forest structure as described by the surveyors was limited to the three broad categories of “heavy timber”, “scattering timber” and “open” of which the latter two may be reliable descriptions but the foremost is less confident. Instead of being used as a reference to stand density, prior analysis has suggested that “heavy timber” was more likely used in relation to presence of big trees (Powell, 2008). We also attempted to use the tree and undergrowth species listed by surveyors for each line to model forested plant associations to extrapolate the generally accepted percent canopy cover for those associations. To do this, we cross referenced these plant associates with the USFS Field Guide to the Forested Plant Associations of the Westside Central Cascades of Northwest Oregon (McCain, 2002). However, limited species information from the historical surveys led to modeled plant associations with the same or similar percent canopy cover. A variation of the fuzzy indices that included these canopy cover estimates produced maps that were much more skewed towards closed canopy forest. Lastly, both variables relating to soil–type and rating–were also evaluated with box plots. Across the seven soil types, the plots show little variability in relation to degree of canopy closure, and over half of the lines (~60%) don’t have a soil type recorded in the survey. Soil type was excluded because it would require academic training or extensive research in the subject to 30 analyze the relationship with forest canopy cover and the reliability of the survey. The overall pattern of soil rating against the fuzzy index is that ratings 1 and 2 are associated with lower degrees of closed canopy than soil rated 3 or 4 which have tighter interquartile ranges and higher medians. The purpose of soil rating was to evaluate the land for potential development or agricultural use; a soil ranked 1 was supposedly better than a ranked 4 soil (Galatowitsch, 1990; Powell, 2008). Working from the assumption that land with fewer trees is easier to clear and therefore more valuable, the plot supports that trend; however, we felt that it extrapolates too far beyond the data. Table 2: Excluded variables VARIABLE UNITS REASON FOR VISUAL EXCLUSION DBH Inches Surveyor bias in species and age of trees selected for bearing trees (Bourdo, 1956; Noss, 1985; Kronenfield and Wang, 2007). Terrain 1 - Level Insufficient data to conduct thorough analysis. 2 - Sloping 3 - Rolling 4 - Broken 31 5 - Mountainous Stand 1 - Open “Heavy timber” may likely Structure have referred to the perceived value of timber 2 - Scattering rather than the density of trees (Powell, 2008). 3 - Heavy timber Soil Type Clay loam Would require additional outside research to investigate plant and soil Clay loam & relationships rocky Clay & rocky Sandy Sandy loam Sandy & rocky Rocky 32 Soil Rating 1 - 4 ranked Uncertainty about survey reliability. Canopy Tree & Too many unknowns and See Appendix II for full table Closure Understory assumptions to make solid Species classifications. presence The comparison of resurveyed lines supports the general pattern that the river corridor in 1869 had a less closed canopy forest than the higher elevation uplands in 1902. For instance, a line graph comparing the number of bearing trees of an original survey line in 1869 and the resurveyed lines in 1902 shows a general increase in trees (Figure 7). Two lines that are significant outliers from this pattern are 130 and 137; survey line 130 had the same number of bearing trees from 1869 to 1902, and line 137 had one more bearing tree in 1869 than in 1902. These survey lines cross the Middle Fork of the Willamette River which likely explains why the lines have the same number of bearing trees for both survey years. Similarly, a line graph of DBHs shows an overall increase in tree size for re-surveyed lines; the average DBH of trees in 1869 was 9.15 inches which increased to 18.28 inches in 1902. This suggests that the bearing trees selected for 1902 may have been present in 1869 but were too small to be selected at that time. Many past studies have pointed out surveyor bias when selecting for bearing trees, both in species and DBH (Noss, 1985). A look at differences in tree species revealed the presence of more species in 1902 than in 1869; a survey line in 1869 averaged 1.7 species while a line 33 surveyed in 1902 averaged 3.3. However, the lower total number of bearing trees in 1869 is likely a factor in there being fewer tree species included in that survey. Figure 6: Comparison of bearing trees of resurveyed lines Figure 7: Comparison of DBH of resurveyed lines 34 Figure 8: Comparison of number of different tree species on resurveyed lines 35 Discussion Our fuzzy set of closed canopy forest structure showed an overall positive relationship between elevation and degree of closed canopy forest; lower elevations along the river course are associated with lower degrees of membership to closed canopy forest, while higher elevations in the uplands support survey lines with higher degrees of membership to closed canopy forest. Because of limitations and bias inherent in the surveyors’ work, using historical land survey data to explore the broader pattern of closed canopy forest was more appropriate than classifying survey lines into discrete vegetation types. While the historical land surveys do provide more information beyond the two variables we used for the fuzzy set, our analysis suggested low confidence in the reliability and relevance of the excluded variables in terms of how they relate to forest canopy structure. Our investigation suggests that on their own, GLO survey data is insufficient in quality or reliability to confidently reconstruct historical forest structure at a spatial grain needed to inform management plans. Comparing Christy and Alverson’s Willamette Valley Historical Vegetation Map As described above, Christy and Alverson (2011) published a historical vegetation map of the Willamette Valley that features discrete vegetation type boundaries. The upper Middle Fork portion of the map does not appear in the publication but is available on-line in the ArcMap shapefile Version 2001_04 (Christy, et. al., 2011). The map has been used to evaluate restoration schemes for seasonal wetlands (Pfeifer-Meister, et. al., 2018), river corridors (Gregory et. al., 2019; Wohl, et. al., 2021), and prairie ecosystems (Dunwiddie and Bakker, 2011) throughout the valley. The authors supplemented the historical land surveys with aerial photographs from the 1930s (that post-date the period they claim their map represents), plat maps, and soil surveys to better inform their estimations of vegetation boundaries. Despite the difference in methodologies, 36 there is some consistency between our fuzzy set for closed canopy forest and Christy et. al.’s map; as Figure 11 shows, sections lines with lower degrees of closed canopy forest roughly overlap with polygons classified as savanna (blue), woodland (purple), and prairie (light orange). Section line 189 is an interesting example of how our methodologies differ. In the survey notes, a prairie is recorded at 30 links along the line, but because our fuzzy set only included the number of bearing trees and distance, the line was classified as a higher degree of closed canopy forest. The section line had 10 bearing trees recorded and the average distance was 33.3 feet which produced the fuzzy “sum” value of 0.85. As Figure 9 shows, line 189 passes through a small portion of the estimated prairie area, which Christy et. al., likely identified with the aerial photographs. Our fuzzy classification is not wrong, but rather represents the difficulty of mapping small patch vegetation within an otherwise contiguous vegetation type. Ways to mitigate this issue in the future would be to conduct fuzzy classification at a finer grain than the section line and perhaps split the line at the quarter section. Alternatively, small patch vegetation could be symbolized separately as point data. 37 Figure 9: Discrete versus fuzzy classification Implications for Future Use of GLO Survey Records Use of GLO surveys to reconstruct historical forest conditions has been done since before surveys were concluded in the 1930s (Sears, 1925), and since then, numerous studies have adapted their own methodologies to account for data limitations over the decades. Now with hundreds of maps in circulation of reconstructions based on GLO surveys, and the increasing need to manage our forests for resiliency, it is imperative to evaluate the credibility of GLO survey records to be used for reconstruction needs. Our results agree with many previous researchers who have concluded that GLOs are too valuable to put aside, but as a standalone source they may be insufficient (Grimm 1984; Manies, et. al., 2001; Kronenfeld and Wang, 2007; Horne, 2010). Bias is inherent in the survey data because of the systematic way surveyors were required to choose and mark bearing trees; while it provided consistency, the resulting data 38 on bearing trees is a nonrandom sample which must be accounted for during analysis (Levine, et. al., 2017). Researchers’ intent to use the surveys for historical vegetation reconstructions differs from the surveys’ original intent to inventory land for sale and settlement. This creates conflict between how researchers want to use the data and limitations on how the data can be used. We suggest that GLO survey data should be used only in conjunction with other sources of information (Schulte and Mladenoff, 2001). Limitations aside, there are simply no other records of historical land conditions like GLOs; as Egan and Howell put it, they are “unexcelled as a source of baseline information” (2001). A key word is “baseline”; with all the data use challenges just discussed, historical land survey data is best suited to complement other data, or to make generalized conclusions. Nonetheless, historical land surveys can be an excellent foundation for fire ecology studies and archaeological investigations. Surveyor instructions from 1850 specifically outlined that surveyors were to record features such as Native American trails, evidence of past fire, windfalls, etc., with location information (Bourdo, 1956), which provide valuable point locations from where researchers can begin their data collection. Plat maps created in association with survey notes were a key historical reference for Pachin et. al.’s (2018) work to reconstruct the locations of Native American settlements on the northwest coast of California, and numerous fire ecology studies have been conducted in conjunction with GLO data (Agee, 1993; Boyd, 1999; Storm 2002). These studies built upon the survey information with other sources in acknowledgement of the surveys’ incomplete and flawed nature. Recommendations for Further Analysis Despite our cautionary approach to working with the GLO survey data by creating a fuzzy classification, further analysis could be conducted to strengthen our fuzzy set of closed 39 canopy forest structure. Since the resulting fuzzy set does show a positive relationship between degree of closed canopy and elevation, an analysis of topographic variables including elevation, slope, and aspect could provide more detail (Thomas-Van Gundy and Nowacki, 2013). Statistical regression for each topographic variable with the fuzzy “sum” for closed canopy forest could provide insight towards how the environmental factors influence the spatial distribution of vegetation. Additionally, analysis of under- and overstory species could be continued with a different approach. Individual species maps could be plotted and compared visually to evaluate species associations; statistical regressions could be run with topographical variables as well to further the investigation between site and species relationships (Dryer, 2001). Although we did not include the percent canopy closure analysis into the final fuzzy set, that analysis supplemented with outside sources could improve the fuzzy set. For example, fire frequency data, or presence of large trees known to develop in more open structure landscapes such as oak and ponderosa pine. In that vein, any further analysis should strongly consider the use of additional sources because of the fine gradient of the study area; on their own, historical land surveys are limited to discerning broad patterns (Schulte and Mladenoff, 2001). 40 Conclusion Despite the broad use of GLO survey data in research, our investigation indicates that the reliability of GLOs as a source depends heavily on the methodology. Fuzzy classification reduces research bias through iterative evaluation and assignment of input variables to degree of membership to closed canopy forest structure. Historical land surveys can be useful for evaluating general trends and patterns as we have shown with our fuzzy “sum” of closed canopy forest structure; however, the reliability and confidence in the accuracy of classification disintegrates as the resolution gets finer grained. The fuzzy “sum” for closed canopy forest could be improved upon through further analysis into topography characteristics, under- and overstory species and fire frequency, but overall represents a carefully evaluated and informed approach to visualizing Contact Era forest structure based on GLO survey notes. 41 Appendix I TERM DEFINITION Chain unit of measurement used by surveyors; one chain consists of 100 links and is a total of 66 feet long DBH diameter in breast height; standardized metric for measuring around tree trunks Distance measurement in chains of the space between a corner or quarter section marker and the bearing tree, or a point on the line and a topographical or cultural feature Fuzzy Overlay m inimum of the fuzzy memberships from the input method: And Fuzzy Overlay maximum of the fuzzy memberships from the input method: Or Fuzzy Overlay increasive function. When the combination of multiple evidence is more method: Sum important or larger than any of the inputs alone Fuzzy Set Theory a way to categorize data or objects based on a gradient to communicate some uncertainty or ambiguity Fuzzy Variable a variable that communicates uncertainty or imprecise boundaries; can be related to linguistic rather than numerical value Link unit of measurement used by surveyors; 100 links makes one chain Normalization statistical process of transforming data from different scales into one common scale Soil Description semantic descriptor of soil materials, i.e., “clay” or “rocky” Soil Rating ranking system 1-4 to evaluate the land for agricultural or settlement potential; 1 indicates good quality soil, and 4 is poor Stand Structure semantic descriptor of vegetation type, i.e., “heavy timber” or “savanna” Terrain semantic descriptor of the land, i.e., “sloping” or “mountainous” Township and rectangular grid system used to systematically survey land; one township is Range 36 square miles or 6 miles by 6 miles. Range denotes west or east of the meridian, i.e., T24SR3E (township 24 south range 3 east) is the third township east of the meridian 42 Willamette true north/south line running from Oregon’s southern border to Canada Meridian Witness a tree selected by a surveyor to mark control points along the section lines Tree/Bearing Tree such as the midpoint or corner. The tree is marked with the township, range, and section number 43 Appendix II Line Tree Understory Guidebook Plant Association (s) % ID Species Species Canopy Cover 188 Fir, cedar Vine maple, Grand Fir/Vine maple; Douglas 63, 67 young firs & Fir/Western Hemlock hemlock 189 Pine, larch Salal & Silver Fir/Western 62, 63, & fir huckleberry Hemlock/Rhododendron/ Salal; Grand 72 Fir/Vine maple; Western Hemlock/ Vine maple/Salal 190 Fir, Vine maple, Grand Fir/Vine maple; Western Hemlock/ 63, 72 hemlock huckleberry & Vine maple/Salal salal 191 Fir, Vine maple, Grand Fir/Vine maple; Western Hemlock/ 63, 72 hemlock huckleberry, Vine maple/Salal young firs & hemlock 208 Fir, Salal & vine Grand Fir/Vine maple; Western Hemlock/ 63, 72 hemlock maple Vine maple/Salal & pine 209 Fir, Rhododendron Western Hemlock/ Rhododendron/Salal 73 hemlock & cedar 211 Fir, Rhododendron, Silver Fir/Western 62, 70 hemlock dogwood, Hemlock/Rhododendron/ Salal; Silver Fir/ hemlock & Rhododendron/Huckleberry/Dogwood buckbrush 44 212 Fir, cedar, Rhododendron, Silver Fir/ Rhododendron/ 70 pine dogwood, soap Huckleberry/Dogwood brush & hemlock 214 Fir, cedar, Soap Brush, vine Douglas Fir/Western Hemlock; Grand 63, 67 pine maple, hemlock Fir/Vine maple 215 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock dogwood, madrone & hemlock 216 Fir, pine, Vine maple, Silver Fir/Rhododendron/ 70 cedar rhododendron & Huckleberry/Dogwood huckleberry 217 Fir, Rhododendron & Silver Fir/Western 62 hemlock hemlock Hemlock/Rhododendron/ Salal & larch 218 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73, 69 hemlock, alder, soap brush Western Hemlock/ Skunk Cabbage cedar & hemlock 219 Fir, Soap brush, alder, Western Hemlock/ Skunk Cabbage 69 hemlock, fallen trees & cedar limbs 220 Fir, pine, Rhododendron, Douglas Fir/Western Hemlock 67 cedar & soap brush & hemlock hemlock with much fallen timber & debris 221 Fir, cedar Rhododendron, Western Hemlock/Dwarf Oregon Grape 74 & arrowwood, soap hemlock brush & hemlock 222 Fir, cedar, Snowbrush, soap Douglas Fir/Poison Oak 70 hemlock brush, hemlock 45 223 Fir, cedar, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 pine & vine maple, hemlock hemlock 224 Fir, Arrowwood, Western Hemlock /Rhododendron/Bear 72 hemlock, hemlock grass cedar 225 Fir, pine, vine maple, Western Hemlock/Vine Maple 78 hemlock, hemlock cedar 226 Fir, pine, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock, vine maple, cedar hemlock 227 Fir, Vine maple, Western Hemlock /Rhododendron/Bear 72 hemlock, hemlock grass cedar 228 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock, soap brush, vine cedar maple, hemlock 229 Fir, pine, Soap Brush, vine Grand Fir/Vine maple; Douglas 63,67 cedar maple & broken Fir/Western Hemlock limbs 230 Fir, Soap Brush, Western Hemlock/ Rhododendron/Salal 73 hemlock, rhododendron, cedar hemlock 231 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock, soap brush cedar 232 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock soap brush 233 Fir, pine, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock, soap brush larch & white fir 46 234 Fir, Rhododendron, Silver Fir/Western 62 hemlock, soap brush, Hemlock/Rhododendron/ larch hemlock Salal 235 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock debris from dead trees 236 Fir, Rhododendron, Western Hemlock / 72 hemlock, vine maple, Rhododendron/Bear grass cedar hemlock 237 Fir, Rhododendron, Silver Fir/ Rhododendron/ Dwarf Oregon 73 hemlock, vine maple, soap Grape pine brush 238 Fir, cedar, Rhododendron, Grand Fir/Dwarf Oregon Grape 73 pine vine maple, hazel 239 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73,73,72 hemlock, vine maple, hazel Grand Fir/Dwarf Oregon Grape; Western cedar Hemlock /Rhododendron/Bear Grass 240 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73, 73 hemlock, vine maple, hazel Grand Fir/Dwarf Oregon Grape larch & & arrowwood pine 241 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock, hazel, vine maple larch & hemlock 242 Fir, larch, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73, 62 hemlock hemlock Silver Fir/Western Hemlock/ Rhododendron/Salal 47 243 Fir, larch, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73, 62 hemlock hemlock Silver Fir/Western Hemlock/Rhododendron/ Salal 244 Fir, larch, Rhododendron, Western Hemlock/ Rhododendron/Salal; 73, 62 hemlock hemlock Silver Fir/Western Hemlock/Rhododendron/ Salal 252 Fir, larch Larch, fir & Douglas Fir/Western Hemlock 67 & hemlock hemlock 272 Fir, pine Rhododendron, Silver Fir/ Rhododendron/ Dwarf Oregon 73 soap brush, Grape hemlock 273 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock soap brush & & pine hemlock 274 Fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 hemlock soap brush & & cedar hemlock 276 Fir, cedar Hazel, vine maple, Western Hemlock/ Rhododendron/Salal; 73,73 & hemlock & Grand Fir/ Dwarf Oregon Grape hemlock soapbush 277 Fir, larch Rhododendron, Western Hemlock/ Rhododendron/Salal 73 & soap brush & hemlock hemlock 278 Young fir, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 larch & vine maple & hemlock hemlock 279 Fir, larch Vine maple, soap Douglas Fir/Western Hemlock 67 & brush & hemlock hemlock 48 280 Young fir, Rhododendron & Western Hemlock/ Rhododendron/Salal 73 pine & hemlock hemlock 281 Fir, larch Rhododendron, Western Hemlock/ Rhododendron/Salal 73 & vine maple, hazel hemlock & hemlock 282 Fir, larch Rhododendron & Western Hemlock/ Rhododendron/Salal 73 & vine maple hemlock 283 Fir, cedar Vine maple, soap Western Hemlock /Rhododendron/Bear 72 & brush, fallen limbs Grass hemlock & trees 284 Fir, Vine maple & Western Hemlock /Rhododendron/Bear 72, 73 hemlock hemlock Grass; Grand Fir/ Dwarf Oregon Grape & cedar 285 Fir, Vine maple & Western Hemlock /Rhododendron/Bear 72, 73 hemlock hemlock Grass; Grand Fir/ Dwarf Oregon Grape & cedar 286 Fir, Vine maple, Western Hemlock /Rhododendron/Bear 72, 73 hemlock hemlock Grass; Grand Fir/ Dwarf Oregon Grape 287 Fir, cedar, Vine maple, Western Hemlock/ Rhododendron/Salal; 73, 72 hemlock rhododendron & Western Hemlock /Rhododendron/Bear hemlock Grass 288 Fir, cedar, Vine maple, soap Western Hemlock /Rhododendron/Bear 72 hemlock brush & hemlock Grass 49 289 Fir, Vine maple, soap Douglas Fir/Western Hemlock 67 hemlock brush & debris from dead trees 290 Fir, vine maple, hazel Douglas Fir/Western Hemlock; Grand Fir/ 67, 73 hemlock hemlock Dwarf Oregon Grape 291 Fir, larch Hemlock, Douglas Fir/Western Hemlock 67 & rhododendron, hemlock honey laurel, snowbush 292 Fir, larch, Vine maple, Western Hemlock /Rhododendron/Bear 72, 73 cedar & hemlock Grass; Grand Fir/ Dwarf Oregon Grape hemlock 293 Fir, larch not noted — — 294 Fir, pine Vine maple & Silver Fir/ Rhododendron/ Dwarf Oregon 73 & cedar dogwood Grape 295 Fir, pine, Rhododendron, Western Hemlock/ Rhododendron/Salal 73 larch & dogwood, honey cedar laurel & vine maple 296 Fir, cedar Hemlock, vine Silver Fir/ Rhododendron/ Dogwood; 70, 71 & maple & dogwood Silver Fir/ Cool wort Foamflower hemlock 297 Fir, cedar Rhododendron, Western Hemlock/ Rhododendron/Salal 73 & pine hemlock, vine maple & dogwood 298 Young fir, not noted — — cedar & pine 50 341 Fir & Vine maple, Grand Fir/Vine maple; Western 63, 72 hemlock huckleberry Hemlock/Vine maple/Salal 342 Cedar, fir, Vine maple, Grand Fir/Vine maple; Western 63, 72 hemlock young fir, salal Hemlock/Vine maple/Salal 343 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock rhododendron 345 Fir, Vine maple, salal, Grand Fir/Vine maple; Western 63, 72 hemlock, huckleberry Hemlock/Vine maple/Salal cedar 346 Old Vine maple, salal Grand Fir/Vine maple; Western 63, 72 growth fir Hemlock/Vine maple/Salal 347 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock, rhododendron pine 348 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock young fir, hemlock 349 Fir, Vine maple, Western Hemlock /Rhododendron/Bear 72 hemlock, young hemlock, Grass cedar fir 350 Fir, Chaparral, vine Grand Fir/Vine maple; Western 63,72 hemlock maple, salal Hemlock/Vine maple/Salal 351 Fir, cedar Vine maple, Silver Fir/Vine maple; Silver Fir/Dwarf 75,77 huckleberry Oregon Grape 51 352 Fir, pine Fir, pine Silver Fir/ Rhododendron/ Dwarf Oregon 73 Grape 353 Fir, pine Huckleberry, vine Silver Fir/ Rhododendron/ Dwarf Oregon 73 maple Grape 354 Fir, cedar Huckleberry, vine Silver Fir/Vine maple; Silver Fir/Dwarf 75,77 maple Oregon Grape 355 Larch, fir, Huckleberry, vine Grand Fir/Vine maple; Western 63,72,62 pine maple, salal Hemlock/Vine maple/Salal; Silver Fir/Western Hemlock/Rhododendron/ Salal 356 Larch, fir Young fir, larch Silver Fir/ Rhododendron/ Dwarf Oregon 73 Grape 357 Larch, fir, Young larch, fir Silver Fir/ Rhododendron/ Dwarf Oregon 73 cedar Grape 358 Fir, larch, Huckleberry, Grand Fir/Vine maple; Western 63,72,62 cedar salal, vine maple Hemlock/Vine maple/Salal; Silver Fir/Western Hemlock/Rhododendron/ Salal 359 Larch, fir, Young fir, larch Silver Fir/ Rhododendron/ Dwarf Oregon 73 cedar Grape 360 Larch, Young fir, Western Hemlock /Rhododendron/Bear 72,73 hemlock, hemlock Grass; Silver Fir/ Rhododendron/ Dwarf fir, pine Oregon Grape 52 361 Fir, larch, Vine maple, Silver Fir/ Rhododendron/ Dwarf Oregon 73 pine young fir, larch Grape 362 Larch, fir, Vine maple, Grand Fir/Vine maple; Western 63,72 cedar huckleberry & Hemlock/Vine maple/Salal salal 363 Fir, Huckleberry, vine Grand Fir/Vine maple; Western 63,72 hemlock, maple, salal Hemlock/Vine maple/Salal cedar 364 Fir, Vine maple, Western Hemlock /Rhododendron/Bear 72 hemlock young firs, Grass & cedar hemlock 365 Fir, Vine maple, Western Hemlock /Rhododendron/Bear 72 hemlock young fir, Grass hemlock 366 Fir, pine, Rhododendron, Mountain Hemlock/ Manzanita 25 hemlock, manzanita, cedar huckleberry 368 Fir, pine, Vine maple, Mountain Hemlock/ Manzanita 25 hemlock, chaparral brush cedar 370 Fir, Vine maple, Mountain Hemlock/ Manzanita 25 hemlock, chaparral brush pine 371 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock young fir, young hemlock 372 Fir, Vine maple, fir, Western Hemlock/ Rhododendron/Bear 72 hemlock, hemlock Grass cedar 373 Fir, Vine maple, salal Grand Fir/Vine maple; Western 63,72 hemlock, Hemlock/Vine maple/Salal pine 53 374 Fir, Young fir, Douglas Fir/Western Hemlock 67 hemlock hemlock 375 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock rhododendron 376 Fir, Vine maple, salal Grand Fir/Vine maple; Western 63,72 hemlock Hemlock/Vine maple/Salal 377 Fir, pine Vine maple, Grand Fir/ Prince's Pine 73 young fir 378 Fir, pine, Huckleberry, Mountain Hemlock/ Manzanita 25 hemlock manzanita 379 Fir, Vine maple, Grand Fir/Vine maple; Western 63,72 hemlock, huckleberry & Hemlock/Vine maple/Salal cedar salal 380 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock manzanette, rhododendron 381 Fir, Vine maple, Douglas Fir/Western Hemlock 67 hemlock manzanette, rhododendron 382 Fir, cedar, Rhododendron, Mountain Hemlock/ Rhododendron 62 hemlock manzanette, salal 383 Fir, cedar, Rhododendron, Mountain Hemlock/ Rhododendron 62 hemlock manzanette 384 Fir, cedar, Rhododendron, Douglas Fir/Western Hemlock; Mountain 67,62 hemlock, manzanette Hemlock/ Rhododendron pine 385 Fir Chaparral, Mountain Hemlock/ Manzanita 25 manzanette 54 386 Fir, pine Chaparral, Mountain Hemlock/ Manzanita 25 manzanette Note: Lines 293 and 298 were too difficult to classify because of the absence of understory information. 55 Bibliography Anderson, R. C., Jones, S. L., & Swigart, R. (2006). Modifying Distance Methods to Improve Estimates of Historical Tree Density from General Land Office Survey Records. The Journal of the Torrey Botanical Society, 133(3), 449–459. Aubé, M. (2008). The pre-European settlement forest composition of the Miramichi River watershed, New Brunswick, as reconstructed using witness trees from original land surveys. Canadian Journal of Forest Research, 38(5), 1159–1183. https://doi.org/10.1139/X07-202 Beckham, S. D. (1981). The Oregon central military wagon road: A history and reconnaissance. Heritage Research Associates Report, Vol. 1, 10-43 Boag, Peter G. Environment and Experience: Settlement Culture in Nineteenth-Century Oregon. Berkeley: University of California Press, 1992. http://ark.cdlib.org/ark:/13030/ft5z09p09z/ Bolliger, J., Schulte, L. A., Burrows, S. N., Sickley, T. A., & Mladenoff, D. J. (2004). Assessing Ecological Restoration Potentials of Wisconsin (U.S.A.) Using Historical Landscape Reconstructions. Restoration Ecology, 12(1), 124–142. https://doi.org/10.1111/j.1526- 100X.2004.00285.x Bourdo, E. A. (1956). A Review of the General Land Office Survey and of Its Use in Quantitative Studies of Former Forests. Ecology, 37(4), 754–768. https://doi.org/10.2307/1933067 Boyd, R. T. (n.d.). Strategies of Indian Burning in the Willamette Valley. 46. Brown, D. G. (1998). Mapping historical forest types in Baraga County Michigan, USA as fuzzy sets. Plant Ecology, 134(1), 97–111. https://doi.org/10.1023/A:1009796502293 Christy, J.A., E.R. Alverson, M.P. Dougherty, S.C. Kolar, C.W. Alton, S.M. Hawes, L. Ashkenas & P. Minear. 2011. GLO historical vegetation of the Willamette Valley, Oregon, 1851-1910. ArcMap shapefile, Version 2011_04. Oregon Biodiversity Information Center, Portland State University. http://pdx.edu/pnwlamp/glo-historical- vegetation-maps-for-oregon-0 Dunwiddie, P. W., & Bakker, J. D. (2011). The Future of Restoration and Management of Prairie-Oak Ecosystems in the Pacific Northwest. Northwest Science, 85(2), 83–92. https://doi.org/10.3955/046.085.0201 Dyer, J. M. (2001). Using witness trees to assess forest change in southeastern Ohio. Canadian Journal of Forest Research, 31(10), 1708–1718. https://doi.org/10.1139/x01-111 Egan, D., & Howell, E. A. (2001). The Historical Ecology Handbook: A Restorationist’s Guide to Reference Ecosystems. Island Press. 56 Fagin, T. D., & Hoagland, B. W. (2011). Patterns from the past: Modeling Public Land Survey witness tree distributions with weights-of-evidence. Plant Ecology, 212(2), 207–217. Franklin, J. F., & Johnson, K. N. (2012). A Restoration Framework for Federal Forests in the Pacific Northwest. Journal of Forestry, 110(8), 429–439. https://doi.org/10.5849/jof.10- 006 Friedman, S. K., & Reich, P. B. (2005). Regional Legacies of Logging: Departure from Presettlement Forest Conditions in Northern Minnesota. Ecological Applications, 15(2), 726–744. Galatowitsch, S. M. (1990). Using the Original Land Survey Notes to Reconstruct Presettlement Landscapes in the American West. The Great Basin Naturalist, 50(2), 181–191. Gregory, S., Wildman, R., Hulse, D., Ashkenas, L., & Boyer, K. (2019). Historical changes in hydrology, geomorphology, and floodplain vegetation of the Willamette River, Oregon. River Research and Applications, 35(8), 1279–1290. https://doi.org/10.1002/rra.3495 Habeck, J. R. (1961). The original vegetation of the mid-Willamette Valley, Oregon. Northwest Science, 35(2), 65-77. Hadley, K. S. (1999). Forest history and meadow invasion at the Rigdon Meadows Archaeological site, western Cascades, Oregon. Physical Geography, 20(2), 116-133. Hanberry, B. B., Yang, J., Kabrick, J. M., & He, H. S. (2012). Adjusting Forest Density Estimates for Surveyor Bias in Historical Tree Surveys. The American Midland Naturalist, 167(2), 285–306. Hanson, C. (2002). Exploiting Federal Lands: The myth of thinning forests. The San Francisco Chronicle. Accessed 2 May 2023 Hessburg, P. F., Agee, J. K., & Franklin, J. F. (2005). Dry forests and wildland fires of the inland Northwest USA: Contrasting the landscape ecology of the pre-settlement and modern eras. Forest Ecology and Management, 211(1), 117–139. https://doi.org/10.1016/j.foreco.2005.02.016 Horne, A. M. (2010). Cautionary review of GLO land survey data and analyses for limitations and inaccuracies in presettlement vegetetion reconstructions [M.A., University of Arkansas]. Retrieved March 14, 2023, from https://www.proquest.com/docview/607263761/abstract/E9937357CB94659PQ/1 Johnson, K., & Swanson, F. (2023). Historical context of old-growth forests in the Pacific Northwest—- Policy, practices, and competing worldviews. Kimmerer, R. W., & Lake, F. K. (2001). The Role of Indigenous Burning in Land Management. Journal of Forestry, 99(11), 36–41. https://doi.org/10.1093/jof/99.11.36 57 Knight, C. A., Cogbill, C. V., Potts, M. D., Wanket, J. A., & Battles, J. J. (2020). Settlement-era forest structure and composition in the Klamath Mountains: Reconstructing a historical baseline. Ecosphere, 11(9), e03250. https://doi.org/10.1002/ecs2.3250 Kronenfeld, B. J., & Wang, Y.-C. (2007). Accounting for surveyor inconsistency and bias in estimation of tree density from presettlement land survey records. Canadian Journal of Forest Research, 37(11), 2365–2379. https://doi.org/10.1139/X07-068 Larsen, A. J. (1940). Roads and the Settlement of Minnesota. Minnesota History, 21(3), 225– 244. Levine, C. R., Cogbill, C. V., Collins, B. M., Larson, A. J., Lutz, J. A., North, M. P., Restaino, C. M., Safford, H. D., Stephens, S. L., & Battles, J. J. (2017). Evaluating a new method for reconstructing forest conditions from General Land Office survey records. Ecological Applications, 27(5), 1498–1513. https://doi.org/10.1002/eap.1543 Lorimer, C. G. (1977). The Presettlement Forest and Natural Disturbance Cycle of Northeastern Maine. Ecology, 58(1), 139–148. https://doi.org/10.2307/1935115 Lyons, J. K., & Beschta, R. L. (1983). Land use, floods, and channel changes: upper Middle Fork Willamette River, Oregon (1936–1980). Water Resources Research, 19(2), 463-471. Manies, K. L., & Mladenoff, D. J. (2000). Testing methods to produce landscape-scale presettlement vegetation maps from the U.S. public land survey records. Landscape Ecology, 15(8), 741–754. https://doi.org/10.1023/A:1008115200471 Manies, K. L., Mladenoff, D. J., & Nordheim, E. V. (2001). Assessing large-scale surveyor variability in the historic forest data of the original U.S. Public Land Survey. Canadian Journal of Forest Research, 31(10), 1719–1730. https://doi.org/10.1139/x01-108 Noss, R. F. (1985). On Characterizing Presettlement Vegetation: How and Why. Natural Areas Journal, 5(1), 5–19. Olsen, L. M., Dale, V. H., & Foster, T. (2007). Landscape patterns as indicators of ecological change at Fort Benning, Georgia, USA. Landscape and Urban Planning, 79(2), 137–149. https://doi.org/10.1016/j.landurbplan.2006.02.007 Olson, K. R., & Suski, C. D. (2021). Food Security: Impact of European Settlement and Infrastructure on Columbia River Salmon Migration. Open Journal of Soil Science, 11(7), Article 7. https://doi.org/10.4236/ojss.2021.117019 Panich, L. M., Schneider, T. D., & Byram, R. S. (2018). Finding Mid-19th Century Native Settlements: Cartographic and Archaeological Evidence from Central California. Journal of Field Archaeology, 43(2), 152–165. https://doi.org/10.1080/00934690.2017.1416849 Pfeifer-Meister, L., Gayton, L. G., Roy, B. A., Johnson, B. R., & Bridgham, S. D. (2018). Greenhouse gas emissions limited by low nitrogen and carbon availability in natural, 58 restored, and agricultural Oregon seasonal wetlands. PeerJ, 6, e5465. https://doi.org/10.7717/peerj.5465 Powell, D. C. (2008).Using General Land Office Survey Notes to Characterize Historical Vegetation Conditions for the Umatilla National Forest. http://orww.org/Osbornes_Project/Rivers/Umpqua/South/References/Powell_2008.pdf Schulte, L. A., & Mladenoff, D. J. (2001). Their Use and Limitations in Reconstructing Presettlement Vegetation. Journal of Forestry, 6. Sears, P. B. (1921). The Natural Vegetation of Ohio. 3. Sears, P. B. (1925). Vegetation Mapping. Spies, T. A., Long, J. W., Charnley, S., Hessburg, P. F., Marcot, B. G., Reeves, G. H., Lesmeister, D. B., Reilly, M. J., Cerveny, L. K., Stine, P. A., & Raphael, M. G. (2019). Twenty-five years of the Northwest Forest Plan: What have we learned? Frontiers in Ecology and the Environment, 17(9), 511–520. https://doi.org/10.1002/fee.2101 Stewart, L. O. (1935). Public land surveys: History, instructions, methods. Collegiate Press, Inc. https://catalog.hathitrust.org/Record/001513757 Storm, L. E. (2002). Patterns and processes of indigenous burning: How to read landscape signatures of past human practices. Ethnobiology and Biocultural Diversity: Proceedings of the 7th International Congress of Ethnobiology, Athens, Georgia, USA, October 2000, 496–508. Stuckey, R. L. (2009). Contributions of Paul B. Sears to Natural Vegetation Mapping in Ohio. https://kb.osu.edu/handle/1811/52460 Tepley, A. J., Swanson, F. J., & Spies, T. A. (2013). Fire-mediated pathways of stand development in Douglas-fir/western hemlock forests of the Pacific Northwest, USA. Ecology, 94(8), 1729–1743. Thomas, J. W, J. F. Franklin, J. Gordon, and K. N. Johnson. 2006. The Northwest Forest Plan: Origins, components, implementation experience, and suggestions for change. Conservation Biology 20(2): 277-87. Thomas-Van Gundy, M. A., & Nowacki, G. J. (2013). The use of witness trees as pyro- indicators for mapping past fire conditions. Forest Ecology and Management, 304, 333– 344. https://doi.org/10.1016/j.foreco.2013.05.025 Tonsfeldt, W. (1994). Review of History of the Willamette National Forest [Review of Review of History of the Willamette National Forest, by L. Rakestraw & M. Rakestraw]. The Public Historian, 16(2), 89–91. https://doi.org/10.2307/3378808 59 Wang, Y.-C. (2005). Presettlement land survey records of vegetation: Geographic characteristics, quality and modes of analysis. Progress in Physical Geography: Earth and Environment, 29(4), 568–598. https://doi.org/10.1191/0309133305pp463ra White, C. A. (1983). A history of the rectangular survey system. US Department of the Interior, Bureau of Land Management. Whitney, G. G., & DeCant, J. P. (2001). The Historical Ecology Handbook: A Restorationist’s Guide to Reference Ecosystems. Island Press, 147-176 Wohl, E., Castro, J., Cluer, B., Merritts, D., Powers, P., Staab, B., & Thorne, C. (2021). Rediscovering, Reevaluating, and Restoring Lost River-Wetland Corridors. Frontiers in Earth Science, 9. https://www.frontiersin.org/articles/10.3389/feart.2021.653623 60