Plant Community Classification and Mapping


Plant Community Classification and Mapping at the Idaho National Laboratory Site 2010

Accurate classification and mapping of vegetation communities have become increasingly important tools for conservation management. By understanding the distribution and condition of plant communities on a landscape, a number of conservation goals can more easily be met including:

  • Determining which community types are intrinsically rare or have been severely degraded
    • Identifying the best remaining occurrences of natural communities across their geographic ranges

  • Assessing the impacts of various land-use scenarios on areas supporting different vegetation types

  • Developing habitat suitability models for predicting species occurrences

  • Ranking vegetation classes with respect to their importance in conservation management planning.

Previous vegetation maps of the INL Site are inadequate to serve these conservation management planning goals, in part, because they are outdated. The most recent effort was almost twenty years ago and does not capture important habitat changes that have occurred since that time including fires, sagebrush die-off and invasion by non-native plants. Prior mapping efforts also lack assessments of accuracy, making it difficult to quantify uncertainty associated with habitat models derived from data contained in those maps. Furthermore, methodologies for vegetation classification and mapping have been refined and standardized since earlier INL Site maps allowing for continuity between classifications and mapping on the INL Site and on neighboring lands managed by other agencies.

Understanding the distribution and condition of plant communities on the INL Site will support the CMP through habitat mapping, development of Habitat Suitability Indices and will help to focus surveys for sensitive species. Additional benefits to land management at the INL Site include guiding revegetation and weed management efforts, increasing the efficiency of assessing environmental impacts and siting plots for inventory and monitoring activities. It will also serve as an important background database for research on the NERP.



The goal of the vegetation community classification and mapping project is to develop an updated vegetation map detailing the distribution of plant communities on the INL Site. Specific objectives are to:

  • Characterize the vegetation community types present on the INL Site

  • Define the spatial distribution of those community types

  • Conduct an accuracy assessment of the resulting map.


Our approach is based on a process developed by the U.S. Geological Survey (USGS), and National Park Service (NPS) for use in land management planning (USDI, NPS 2009) and includes two parallel tasks, plant community classification and map unit delineation. Plant community classification entails multivariate analysis of applicable historical vegetation data sets and a current project-specific vegetation data set, resulting in a statistically definable list of vegetation classes that can be reconciled with U.S. National Vegetation Classification System (NVCS)-defined vegetation associations (FGDC 2008). The map unit delineation process consists of generating polygons using current digital color-infrared aerial imagery, several ancillary data layers, and image processing techniques to define areas of similarity or dissimilarity across the INL Site. Products of these efforts are then reconciled by assigning vegetation classes to map units resulting in a map for which accuracy will then be assessed.

Accomplishments through 2009

Community Classification

The plant community classification was refined in 2009. Some of the clusters resulting from the classification model selected based on “best fit” to the INL Site vegetation data had high beta diversity scores relative to average beta diversity for all of the clusters. These clusters were reanalyzed using methods similar to those employed for the primary classification effort. The secondary analysis resulted in further splitting a few of the clusters. Two additional clusters were split based on the presence of indicator species that were not necessarily abundant enough or present in enough of the sampled plots to emerge as a distinct cluster in the primary classification analysis or to be reanalyzed in a secondary analysis, but clearly suggest a unique plant community. A total working class list of 27 plant communities were identified and named according to NVCS conventions. A dichotomous key was developed based on the working class list to facilitate rapid characterization of plant communities in the field and was used for collecting data to support accuracy assessments.

Field Sampling

During the summer of 2009, two field crews collected vegetation community data needed for independent validation of the final map polygons. We used a constrained random sampling design to select the validation plot locations. We buffered all roads and major facilities by 100 m and 1100 m using a GIS, which resulted in 1 km of potential sampling area centered on access roads. We also designated a minimum of 500 m between randomly selected plots to avoid sampling plots in close proximity to each other.

Initially, we randomly selected 500 points across the INL Site. Each point location was reviewed and any plot that fell directly on or near obvious boundaries (e.g., wildland fire boundaries) was shifted into the adjacent polygon so the edge was not included in the plot. Eighteen points were removed because they fell on steep talus slopes that would be dangerous for field crews, were located within fenced areas near facilities with restricted access, or fell in heterogeneous areas of the INL Site that contained multiple distinct boundaries (e.g., canals and roads both present). We did not want to neglect sampling the areas outside of the 1100 m road buffer, so we manually selected twenty additional plots that fell in remote regions of the INL Site. Near the end of the field season we evaluated which vegetation communities had the lowest sample sizes and made an effort to target those communities during the last few weeks. We guided field crews to search specific regions where those communities are known or expected to be found, or allowed the crews to opportunistically select a plot location when any rare community was found. We sampled a total of 534 validation plots in 2009



Each validation plot consisted of a sampling plot array which included a focal plot and four peripheral subplots in the cardinal directions. The sampling array is intended to provide information about vegetation communities across a larger extent that more closely represents the scale communities are being delineated on the imagery, and to encompass some of the vegetation mixtures that occur across the INL Site. At each validation subplot, we collected Global Positioning System (GPS) data, created a complete species list and assigned each species a categorical ranking of abundance, and each subplot was assigned to a vegetation community using the field key generated from the 2008 statistical results. We also took representative photographs from the focal plot in all four cardinal directions and additional photographs on the outer edge of each peripheral plot aimed toward the focal plot.

Image Delineations
Draft polygon delineations were completed for the INL Site in 2009. The delineation process relied on a number of datasets to identify and map community boundaries, such as the 2007 INL Site color-infrared imagery, the 2004 Idaho National Agricultural Imagery Program (NAIP) color imagery, and various GIS datasets. Once the delineations were completed, the next step was to start assigning community labels to each polygon in the map. In the fall of 2009, we drove many of the roads across the INL Site visiting observation points to identify what communities were within the map polygons. The observation points were manually selected according to draft map polygon boundaries and are independent of the validation points that will be used to calculate class and overall map accuracies. After visiting numerous locations on the ground it became apparent that the first draft delineations were too detailed and needed to be combined and collapsed with adjacent polygons. We also learned that the patterns evident in the imagery do not always correspond to vegetation community patterns, and in some cases regions that look very different in the imagery appear the same on the ground, or areas that look similar in the imagery actually exhibit greater variation on the ground. We also observed that in most cases communities on the INL Site can be described as mixtures of the communities we statistically defined, and generally at least two communities are present at the 1:12,000 mapping scale used to create the map. Using the ground observation data we revised the initial delineations where appropriate and began assigning community class labels to all polygons in the map.

Because the map accuracy assessment has not been completed, we do not have any results to report at this time.

Plans for Continuation
In 2010, the plant community classifications will be completed and a list of plant communities occurring on the INL Site will be finalized and cross-walked to the NVCS. An updated key to plant communities in the INL Site will also be generated. Plant community descriptions will be written to accompany the final map and we will utilize a format similar to that being used by the NPS at the Upper Columbia Basin Network of Parks (including Craters of the Moon National Monument and Preserve).

The final vegetation community map will be completed in spring 2010, and the accuracy assessment stage will follow. We plan to calculate numerous accuracy metrics to accompany the map so users with different intended applications can best understand the utility and possible limitations of the map.

The project and final report will be completed in 2010.

2009 Plant Classification and Mapping Report

2010 Plant Classification and Mapping Report

2011 Final Plant Classification and Mapping Report