APPENDIX 1.
Historical background
Prior to European settlement, the North American landscape was already altered by Native American activities. Charcoal, isotope, and archeological evidence suggest that tribes along the East Coast used fire to clear forests long before European settlers arrived (Douglas, 1994; Springer et al., 2010). These activities created small but noticeable changes in the pollen record, including decreased contributions from forest taxa and increased representation from crop, disturbance, and early successional species (Baker et al., 1993; Black et al., 2006). Additionally, deposits in New Jersey and Pennsylvania indicate that Native American maize agriculture caused decreased forest cover and increased sedimentation in valley-bottoms between 1000 and 1600 C.E. (Stinchcomb et al., 2011). Therefore, floodplains were both hydrologically and ecologically altered by prehistoric peoples (Asch Sidell, 2008; Stinchcomb et al., 2011); however, the magnitude and impact of these alterations were significantly less than those caused by European settlers during the seventeenth and eighteenth centuries.
European settlement significantly affected the hydrogeomorphology and associated wetland and riparian flora throughout the Piedmont region due, in large part, to the mills, forges, furnaces, and mining operations settlers built to support their communities (Walter and Merritts, 2008; Voli et al., 2009; Wegmann et al., 2012). These structures were dependent on hydropower provided by tens of thousands of dams erected on streams throughout the region (Walter and Merritts, 2008; Merritts et al., 2013). In addition to this expansive infrastructure, European colonists transformed hundreds of thousands of acres of upland, old-growth forest into agricultural fields in the northeastern United States (Costa, 1975; Thompson et al., 2013). Many of the crops planted, such as tobacco, were harmful to soil quality, so settlers frequently cleared more land as fields were exhausted and abandoned after a few seasons (Gottschalk, 1945; Middleton, 1953).
These agricultural clearing and tilling practices triggered large-scale topsoil erosion throughout the Piedmont region (Costa, 1975; James, 2013; Merritts et al., 2013). The fine-grained, extremely mobile, phosphorus and nitrogen-rich silts deposited in valley-bottoms and floodplains from colonial activities are collectively referred to as “legacy sediments” (Walter and Merritts, 2008; James, 2013).
Dams built across the valley-bottoms decreased flow velocity by up to 60%, causing the legacy sediments to settle out of the water column and to form alluvial deposits behind milldams that could stretch hundreds of meters, often upstream to the next dam (Walter and Merritts, 2008; Merritts et al., 2010b; Merritts et al., 2013). In the Pennsylvania Piedmont, two-thirds of this sediment was eventually deposited on modified floodplains and on lower hill slopes within river valleys, causing several meters of floodplain elevation relative to hydrologic base level along the entire lengths of streams and across the entire widths of valley floors (Walter and Merritts, 2008; Hartranft et al., 2011; James, 2013; Wegmann et al., 2013; Weitzman et al., 2014).
As alternative energy sources became available, dams were either removed or abandoned and eventually breached (Merritts et al., 2013). These events triggered stream incision through the reservoir sediments, ultimately forming meandering streams with elevated banks and characteristic water flow patterns involving fast-flowing riffles with artificially steep gradients at dam sites, and pools where ponds had formed behind the milldams (Simon, 1989; Walter and Merritts, 2008; Merritts et al., 2010b). This process was repeated tens of thousands of times across the landscape, dramatically shifting the hydrogeomorphology of first to third order streams from small, vegetated, anabranching channels that were well-connected with the water table and their floodplains to the single-channel, steep-banked streams seen today (Simon, 1989; Walter and Merritts, 2008; Voli et al., 2009; Hartranft et al., 2011).
APPENDIX 2.
Environmental restoration
Increasingly accurate models of the regional pre-settlement wetland and riparian zone flora, especially including trees, are critical for guiding restoration and conservation decisions to ensure a resulting, functioning ecosystem (Weisberg et al., 2013; Johnson, 2014; Smucker and Detenbeck, 2014). Communities assembled from Holocene floral reconstructions can be carefully used as baseline targets for restoration and conservation projects in areas influenced by human activities (Palmer et al., 2005; Bennion et al., 2011).
Post-milldam streams with pools, riffles, and point bars (Walter and Merritts, 2008) were once considered “natural”, and they were commonly used as models for environmental restoration projects. However, the experimental “restorations” continued to modify riparian zone vegetation and prevent streams from breaching onto their floodplains (FISRWG, 2008), resulting in limited overall success (Gutshall and Oberholtzer, 2011; Hartranft et al., 2011). Recently, it has been demonstrated that a more sustainable approach is to completely remove legacy sediments, then restore the naturally occurring riparian vegetation at the original base level wherever possible (Voli et al., 2009; Gutshall and Oberholtzer 2011; Hartranft et al., 2011; Niemitz et al., 2013). Rapid planting of native flora, including hardwood trees, following legacy sediment removal is necessary because weedy species are more likely to colonize the disturbed environment first, excluding the native riparian vegetation and decreasing ecosystem function and riparian zone restoration success (Shafroth et al., 2002; Stanley and Doyle, 2003; Doyle et al., 2005; Gutshall and Oberholtzer, 2011; Kaase and Katz, 2012). More than three decades may be needed for hardwood species to recolonize an altered area after dam removal (Stephens, 2014; Kim et al., 2015), and vegetation along the riparian zone is one of the most important drivers of restoration success (FISRWG, 2008; Gutshall and Oberholtzer, 2011).
Successful restoration, including reestablishing a stream’s access to its floodplain, reducing erosion, and creating a self-sustaining wetland and riparian zone, necessitates an understanding of the pre-European settlement floral community. After removal of legacy sediments, future restoration efforts that plant documented prehistoric wetland and lowland species, including those described in this study, may be able to recreate native and ecologically functional riparian and valley-margin forests that closely mimic the unaltered environment. Additionally, restoration projects that include legacy sediment removal and the re-establishment of native species in the riparian zone can decrease the amount of sediment eroded and transported to downstream areas like the Chesapeake Bay (Correll et al., 1992; Walter et al., 2007; Gutshall and Oberholtzer, 2011).
APPENDIX 3.
Advantages of subfossil leaves
Most Holocene floral reconstructions are based on fossil pollen because it is often abundant and well preserved (Gajewski, 1988; Dull, 1999). However, pollen is susceptible to temporal and spatial averaging, and it is frequently only identifiable to the genus level (Birks and Birks, 2000). Fruits, seeds, and woods are also regularly used, but they are likewise prone to temporal and spatial averaging and are frequently not identifiable to the species level. On the other hand, Holocene leaves are usually identifiable to the species level and represent the most temporally and spatially localized information because they cannot be reworked. However, leaf subfossils are not often used because they are fragile, difficult to process, and quite rare in comparison to more durable fruits, seeds, woods, and palynomorphs (Goetcheus and Birks, 2001). Fruit, seed, and pollen assemblages are generally considered to have temporal resolutions of 
years because they are highly resistant to decomposition and abrasion and are therefore prone to reworking, redeposition, and temporal averaging (Behrensmeyer et al., 2000; Greenwood, 1991). In contrast, leaf floras are considered to have higher temporal resolution, generally between 
years (Behrensmeyer et al., 2000; Birks and Birks, 2000). Additionally, leaf macrofossils frequently represent a more local community than either fossil pollen or seeds because they are transported shorter distances (Wing and DiMichele, 1995; Davies-Vollum and Wing, 1998; Behrensmeyer et al., 2000; Birks and Birks, 2000). When combined with published paleontological data from seeds and pollen, community reconstructions based on leaf macrofossils can overcome many of the spatial and temporal disadvantages associated with pollen and seeds alone, providing a more accurate view of ancient communities (Birks and Birks, 2000).
