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. All of the interpreted and remotely sensed data were converted to Geographic Information System (GIS) databases using ArcMap© software and are included in a comprehensive geodatabase. Draft maps created from the vegetation classification were field-tested and revised before independent ecologists conducted an assessment of the map’s accuracy during 2014.

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. Mapping began in earnest in 2007 by re-sampling the 2002 imagery to a 3-meter pixel resolution and then segmenting it using eCognition software. Initial segments were created to delineate obvious landforms (e.g. open water and fields) and physiognomic features (e.g. grasslands versus woodlands). Following segmentation, the lines were exported as ArcInfo shapefiles and converted to ArcInfo coverages. The resulting coverages were run through a series of smoothing routines until no obvious artificial or relict breaks in the lines were visible. Following smoothing, the line-work was manually cleaned to remove extraneous lines, small polygons, and polygons that obviously split a homogenous stand of vegetation.

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 digital vegetation map for GARI was developed as a personal geodatabase using Environmental Systems Research Institute (ESRI) ArcGIS software. The geodatabase includes a point feature class for locations of plots and two polygon-feature classes (clipped by the park boundary and unclipped) for vegetation, including non-vegetated land cover. The vegetation map includes 31 map classes. Upland communities comprise about 86% of the park area and are represented by 13 map classes. Two upland map classes each include patches of two associations, all others represent single associations.

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. The TOP 2015 imagery was mosaiced and manipulated using image processing and segmentation techniques (e.g. unsupervised image classification, normalized difference vegetation index, etc.) to highlight any subtle vegetation signature differences. All of the preliminary results were evaluated for usefulness and the best examples were first converted to digital lines and polygons, were next combined with other relevant AMIS GIS layers (such as the roads network), and the results were used as the base layer for the new AMIS vegetation mapping effort. Building off the base layer, all relevant lines and polygons were exported as shapefiles and converted to ArcGIS coverages. The resulting coverages were run through a series of smoothing routines provided in the ArcGIS software. Following the smoothing, all digital line-work was manipulated to remove extraneous lines, eliminate small polygons, and merged polygons that split obvious stands of homogeneous vegetation. The cleaning stage was considered complete when all resulting polygons matched homogenous stands of vegetation apparent on the TOP 2015 imagery. At this point, the mapping shifted to manual techniques and all vegetation lines and polygons were visually inspected and manually moved, edited and/or updated as needed.

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. Our final map product is a geographic information system (GIS) database of vegetation structure and composition across the Crater Lake National Park terrestrial landscape, including wetlands. The database includes photos we took at all relevé, validation, and accuracy assessment plots, as well as the plots that were done in the previous wetlands inventory. We conducted an accuracy assessment of the map by evaluating 698 stratified random accuracy assessment plots throughout the project area. We intersected these field data with the vegetation map, resulting in an overall thematic accuracy of 86.2 %. The accuracy of the Cliff, Scree & Rock Vegetation map unit was difficult to assess, as only 9% of this vegetation type was available for sampling due to lack of access. In addition, fires that occurred during the 2017 accuracy assessment field season affected our sample design and may have had a small influence on the accuracy. Our geodatabase contains the locations where particular associations are found at 600 relevé plots, 698 accuracy assessment plots, and 803 validation plots.

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. Following the vegetation plot data analysis, the preliminary vegetation map was edited and refined to produce a revised preliminary vegetation map prior to thematic accuracy assessment. Using ArcMap 9.2 (ESRI 1999-2006), polygon boundaries were revised on screen based on the plot data, field observations, classification analyses, aerial photography signatures, and topographic maps. Each polygon was assigned the NVC Community Element Global (CEGL) code of a preliminary vegetation association based on the information sources listed above. Second, third, and fourth CEGL code choices were entered in cases of uncertainty, or for polygons representing mosaics of two or more types.

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. The final MNRR vegetation map consists of 3,590 polygons totaling 69,011 ac (27,928 ha). Mean polygon size is 12.1 ac (4.90 ha). Of the total area, 37,789 ac (15,293 ha) or 55% represent natural or ruderal vegetation map classes. Agricultural vegetation such as cultivated crops and pasture account for approximately 6,238 ac (2,524 ha) or 9% of the total mapped area. Non-vegetated barren land was rare, only accounting for 45 ac (18 ha) or 0.1%. Developed areas such as open mowed fields, parking lots, buildings, and others account for approximately 1,562 ac (632 ha) or 2%. Open water is the most widespread land cover class, with an area of approximately 23,422 ac (9,479 ha) or 34% of the total mapped area. Within the total area occupied by vegetation map classes, forest and woodland types were the most extensive (17,007 ac (6,882 ha) or 46%), followed by herbaceous types (11,457 ac (4,636 ha) or 31%), sparse types (6,287 ac (2,544 ha) or 16%), and shrubland types (3,038 ac (1,229 ha) or 7%).

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. We converted the photointerpreted data into a format usable in a geographic information system (GIS) by employing three fundamental processes: (1) orthorectify, (2) digitize, and (3) develop the geodatabase. All digital map automation was projected in Universal Transverse Mercator (UTM), Zone 16, using the North American Datum of 1983 (NAD83). Orthorectify: We orthorectified the interpreted overlays by using OrthoMapper, a softcopy photogrammetric software for GIS. One function of OrthoMapper is to create orthorectified imagery from scanned and unrectified imagery (Image Processing Software, Inc., 2002). The software features a method of visual orientation involving a point-and-click operation that uses existing orthorectified horizontal and vertical base maps. Of primary importance to us, OrthoMapper also has the capability to orthorectify the photointerpreted overlays of each photograph based on the reference information provided. Digitize: To produce a polygon vector layer for use in ArcGIS (Environmental Systems Research Institute [ESRI], Redlands, California), we converted each raster-based image mosaic of orthorectified overlays containing the photointerpreted data into a grid format by using ArcGIS. In ArcGIS, we used the ArcScan extension to trace the raster data and produce ESRI shapefiles. We digitally assigned map-attribute codes (both map-class codes and physiognomic modifier codes) to the polygons and checked the digital data against the photointerpreted overlays for line and attribute consistency. Ultimately, we merged the individual layers into a seamless layer. Geodatabase: At this stage, the map layer has only map-attribute codes assigned to each polygon. To assign meaningful information to each polygon (e.g., map-class names, physiognomic definitions, links to NVCS types), we produced a feature-class table, along with other supportive tables and subsequently related them together via an ArcGIS Geodatabase. This geodatabase also links the map to other feature-class layers produced from this project, including vegetation sample plots, accuracy assessment (AA) sites, aerial photo locations, and project boundary extent. A geodatabase provides access to a variety of interlocking data sets, is expandable, and equips resource managers and researchers with a powerful GIS tool.

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. Several imagery sources were used to derive the vegetation map, primarily Light Imaging and Detection Radar (LiDAR) and aerial imagery from the National Agriculture Imagery Program. The eCognition software package and Berkley Imaging Segmentation was used to create the initial image segments and polygon map. The mean values of variable inputs were summarized for each of the training data polygons and were used to generate a predictive non-parametric model using RandomForest in the statistical program R. The model was then applied to all polygons. The resulting draft map was reviewed by experts familiar with the vegetation types of the area. The final map includes 24 classes, representing 3 land cover types and 21 alliance-based map classes. A field-based, blind random sample accuracy assessment of the map was carried out in mid-August 2011 and December 2011. A total of 175 accuracy assessment points were collected. Based on the assessment, the total map accuracy was 88.9%, exceeding the program standard of 80%.

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. Several imagery sources were used to derive the vegetation map, primarily Light Imaging and Detection Radar (LiDAR) and aerial imagery from the National Agriculture Imagery Program. The eCognition software package and Berkley Imaging Segmentation was used to create the initial image segments and polygon map. The mean values of variable inputs were summarized for each of the training data polygons and were used to generate a predictive non-parametric model using RandomForest in the statistical program R. The model was then applied to all polygons. The resulting draft map was reviewed by experts familiar with the vegetation types of the area. The final map includes 24 classes, representing 3 land cover types and 21 alliance-based map classes. A field-based, blind random sample accuracy assessment of the map was carried out in mid-August 2011 and December 2011. A total of 175 accuracy assessment points were collected. Based on the assessment, the total map accuracy was 88.9%, exceeding the program standard of 80%.