In lieu of a uniform mapping of the Chesapeake and Ohio Canal National Historical Park corridor at very high-resolution using UAS, this study developed a multi-scale workflow, where (1) geospatial modeling methods and (2) historic image analysis were used to constrain the areal extent of (3) detailed field and unmanned aerial systems (UAS) observation. 1_Geospatial Modeling Methods: Harperella habitat characteristics reported by literature sources and corroborated by extremely limited harperella occurrence data (in the form of GPS locations), were compiled into a geospatial prediction model (GPM) to characterize the extent of harperella habitat for the region between Sideling Hill Wildlife Management Area and Harper’s Ferry National Park. 2_Historical Image Analysis: Analysis consisted of visual examination and manual delineation of in-channel bars within the Potomac River and its larger tributaries, including the Cacapon River, lower Tonoloway Creek, Sleepy Creek, lower Lick Run, and Back Creek. This manual delineation was conducted for several dates of historic aerial imagery, including 2009, 2011, 2013, 2016, and 2017. Persistence of different parts of the in-channel bars over time was mapped by intersecting the 5 years of interpretations. 3_UAS image acqusition of AOIs: UAS imagery was collected to facilitate detailed observation, terrain modeling, and documentation of 10 AOIs (shown in figure). National Park Service regulations restrict the take-off and landing of UAS on park property, so the edge of in-channel bars within the Potomac River was used to enable line-of-sight UAS image acquisition covering the entire AOI. The USGS and the Aerial Vision Group, LLC sought permission to fly a small UAS from both the Maryland and West Virginia Department of Natural Resources (DNR). DEM and Orthophoto datasets are available by request - please email jdewitt@usgs.gov

During 2018, uncrewed aerial vehicles (UAVs or 'drones') were used to collect spatially referenced aerial imagery from 20 management units (sites) enrolled in the Phragmites Adaptive Management Framework, a collective learning program developed by the Great Lakes Phragmites Collaborative. Management units were located in Michigan, Ohio, and Wisconsin (USA). Invasive Phragmites australis (hereafter "Phragmites") had been managed at each management units some time previously by the landowner or land manager, and aerial imagery was then collected to create cover classifications distinguishing live and dead Phragmites from the surrounding landscape using object-based image analysis with training based on ground-truth field data and photos. Standard color (RGB) imagery was collected at all 20 management units, and near-infrared (NIR) imagery was collected at 2 of the 20 management units. Accuracy for the classifications was assessed by comparing cover classifications to ground truth data via confusion matrices. The accuracy associated with generating cover classifications by RGB and NIR imagery were also compared.

During 2018, uncrewed aerial vehicles (UAVs or 'drones') were used to collect spatially referenced aerial imagery from 20 management units (sites) enrolled in the Phragmites Adaptive Management Framework, a collective learning program developed by the Great Lakes Phragmites Collaborative. Management units were located in Michigan, Ohio, and Wisconsin (USA). Invasive Phragmites australis (hereafter "Phragmites") had been managed at each management units some time previously by the landowner or land manager, and aerial imagery was then collected to create cover classifications distinguishing live and dead Phragmites from the surrounding landscape using object-based image analysis with training based on ground-truth field data and photos. Standard color (RGB) imagery was collected at all 20 management units, and near-infrared (NIR) imagery was collected at 2 of the 20 management units. Accuracy for the classifications was assessed by comparing cover classifications to ground truth data via confusion matrices. The accuracy associated with generating cover classifications by RGB and NIR imagery were also compared.

The files linked to this reference are the geospatial data created as part of the completion of the baseline vegetation inventory project for the NPS park unit. Current format is ArcGIS (10.6.x) file geodatabase. To map the vegetation and land cover of the parks within the National Capital Region, the region initiated collective mapping efforts at 10 parks (NPS unit codes: ANTI, CATO, CHOH, GWMP, HAFE, MANA, MONO, NACE, PRWI, WOTR). NatureServe assisted with field plots, accuracy assessment, and with building the classification for the vegetation map. This geospatial dataset only covers Chesapeake & Ohio Canal National Historical Park.

The files linked to this reference are the geospatial data created as part of the completion of the baseline vegetation inventory project for the NPS park unit. Current format is ArcGIS file geodatabase but older formats may exist as shapefiles. A total of 88,479 acres (35,806 ha) comprising LAMR, ALFL and its environ was mapped. The area mapped within the Park boundary was 43,037 acres (17, 417 ha). Thirty-four map units were developed to describe the landscape. Of all the map units, the most frequently occurring within the entire mapping area was Map Unit 8, Honey Mesquite Shrubland with 825 polygons ranging in size from under 0.01 acres to over 285 acres. The most abundant map unit in terms of area was Map Unit 17, Upland Slopes/Rolling Hills Vegetation Complex at 27,128 acres or about 31% of the total mapped area but 18% of the Park. Normally the standard minimum mapping unit for NPS vegetation mapping projects is defined as 0.5 hectares. However, this definition was used as a guideline and the actual minimum threshold defined by the high resolution of the aerial photography was more in the range of 1/4 acre.

The files linked to this reference are the geospatial data created as part of the completion of the baseline vegetation inventory project for the NPS park unit. Current format is ArcGIS file geodatabase but older formats may exist as shapefiles. A total of 88,479 acres (35,806 ha) comprising LAMR, ALFL and its environ was mapped. The area mapped within the Park boundary was 43,037 acres (17, 417 ha). Thirty-four map units were developed to describe the landscape. Of all the map units, the most frequently occurring within the entire mapping area was Map Unit 8, Honey Mesquite Shrubland with 825 polygons ranging in size from under 0.01 acres to over 285 acres. The most abundant map unit in terms of area was Map Unit 17, Upland Slopes/Rolling Hills Vegetation Complex at 27,128 acres or about 31% of the total mapped area but 18% of the Park. Normally the standard minimum mapping unit for NPS vegetation mapping projects is defined as 0.5 hectares. However, this definition was used as a guideline and the actual minimum threshold defined by the high resolution of the aerial photography was more in the range of 1/4 acre.

Vegetation and elevation data were collected using a real-time kinematic global positioning system (RTK GPS) in coastal wetlands at the National Park Service’s Timucuan Ecological and Historic Preserve in summer 2021 and winter 2022 (n = 362). For each 0.5-by-0.5 m plot, the following data were collected: 1) percent cover by vegetation species; 2) percent cover by vegetation classes based on height (that is, carpet, herbaceous, and woody); 3) mean height of the dominant species; 4) water depth; 5) marsh class based on dominant vegetation species; and 6) location and elevation (that is, northing, easting, and elevation).

Vegetation and elevation data were collected using a real-time kinematic global positioning system (RTK GPS) in coastal wetlands at the National Park Service’s Timucuan Ecological and Historic Preserve in summer 2021 and winter 2022 (n = 362). For each 0.5-by-0.5 m plot, the following data were collected: 1) percent cover by vegetation species; 2) percent cover by vegetation classes based on height (that is, carpet, herbaceous, and woody); 3) mean height of the dominant species; 4) water depth; 5) marsh class based on dominant vegetation species; and 6) location and elevation (that is, northing, easting, and elevation).

Since 2008, large-scale restoration programs have been implemented along the Willamette River, Oregon, to address historical losses of floodplain habitats for native fish. For much of the Willamette River floodplain, direct enhancement of floodplain habitats through restoration activities is needed because the underlying hydrologic, geomorphic, and vegetation processes that historically created and sustained complex floodplain habitats have been fundamentally altered by dam construction, bank protection, large wood removal, land conversion, and other influences (for example, Hulse and others, 2002; Wallick and others, 2013). An emerging management issue in the Willamette River floodplain and focus of river restoration efforts from 2015-2021 was the treatment of invasive, non-native, mat-forming emergent aquatic macrophyte, water primrose (Ludwigia hexapetala and L. peploides - henceforth Ludwigia). Ludwigia has been widely observed within off-channel features of the Willamette River in northwestern Oregon, raising questions about the effects of this plant on habitat conditions for native fish and wildlife. This document describes repeat digital maps of Ludwigia and other cover classes that were developed using random forest classification in R statistical analysis software of WorldView 2 and 3 satellite images acquired in 2012, 2015, and 2018 in two reaches of the Willamette River near Albany and Salem, Oregon. The Albany reach extends from floodplain kilometer (FPKM) 160 to 138, and the Salem reach extends from FPKM 116 to 91. This mapping effort is intended as preliminary reconnaissance to aid in determining the spatial and temporal patterns of Ludwigia presence and inform more comprehensive, river-scale mapping efforts in the future. The digital maps are rasters representing coverage of Ludwigia and seven other land cover classes produced from random forest classification mapping techniques. Mapping datasets are accompanied by training polygons and accuracy assessments of the classification models.

Plant data were collected on 16, 1-ha experimental blocks in C3 and Marsh Creek units of Seney National Wildlife Refuge, 2006-2010. The percent cover of each plant taxon, moss as a group, and open area was recorded sing a modified line-intercept method for each of 25 sampling points within a block. Sampling points were equidistantly spaced, with 5 points spaced 25 m apart along each of 5 transects also spaced 25 m apart within each sampling block. Within each unit, we selected 4 pairs of blocks representing sedge-shrub habitat, with one of the pair assigned to spring burning (C3, May 2008) or summer burning (Marsh Creek, 2007 and 2008). This before-after-control-impact design provided for data collection two growing seasons before the burn (2006, all 16 blocks; 2007, 15 blocks) and two-three growing seasons (2008, 2009, and 2010; 14 blocks) after burning; the unburned plot of each pair served as the control, and the burned plot of each pair the treatment. Taxa were identified to species where possible,but some taxa were identified only to genus, and some species were merged for analyses because of challenges separating similar species. These data were used in conjunction with environmental data within each block to examine the effects burning on the plant community.