VEGETATION DATA SET

Because INFORMS is designed for NEPA analysis in the Forest Service, one of the problems has been how to look at future conditions of vegetation with and without treatment. This was easily done for single stands using the Forest Vegetation Simulator (FVS) but unless there is a stand exam on each stand it could not be done for the landscape of vegetation data for the analysis area.

The solution to this problem is the use of the Most Similar Neighbor Analysis (MSN) technology. The INFORMS application team has found that by the use of these two technologies (FVS & MSN) integrated with Landsat image data. We can create high quality vegetative data sets to develop current and future conditions using these technologies.

The current evidence for the quality of these vegetation data sets is anecdotal. Several examples of recent stand exams compared to MSN imputed data have been evaluated. These comparisons indicate that this method works quite well in producing accurate representations of vegetation for analysis. These anecdotal analyses have been done on National Forests in the western United States, including Idaho, Oregon, New Mexico, and Colorado.

A standardized method of accuracy assessment is being developed in cooperation with the USFS - Remote Sensing Applications Center. This approach will be pier reviewed and will give more accepted accuracy data for NEPA analysis and other uses of these data sets.

The requirements to support the production of this data set are:

Stand Polygon GIS Layer
Recent Spring Landsat Scene
Stand Exams well distributed on all veg types on 10% or more of the polygons
Stand exam data loaded into FSVeg database to standardized the data source
Digital Elevation Model buffered for the MSN Analysis Area

These are commonly available types of data. The tools within the INFORMS application will grow the stands to the date of the Landsat scene to calibrate them more accurately to the reflectance values represented. These stand exams become what is known as reference stands. Based on the Landsat and DEM data the MSN application will find the most similar stand to use to represent each stand that does not have current data. Then a table is populated called the “for/use” table. This is simply a pointer table that is used when an application in INFORMS needs data for a stand that has no stand exam on it INFORMS looks to the pointer table.

For example when stand_id (2742) has no data the pointer table has a record that says FOR stand_id (2742) USE stand_id (3544). So we now have a condition where we have data effectively in all stands across the landscape because there is a record in the “for/use” table for all stands on the analysis landscape. This is very valuable for project NEPA analysis but that will be addressed in a later section.

It turns out that this also produces an opportunity to create a high quality current vegetation GIS coverage with attributes for all stands. The current output is used to produce fuels , wildlife habitat and other vegetation representations for use in public meetings, for other fuels analysis and for planning fire suppression activities with FARSITE .

INFORMS currently produces more attributes that are useful to all types of resource analysis. These attributes include:

Basal Area of tree species
Tree species of the first 3 species in descending dominate order
Average stands height
Overstory Average Diameter
Plant species total cubic foot volume or biomass

These are attributes that traditional image processing methods usually can not produce. These can be produced to a high degree of accuracy that can be improved with more stand exam sampling if needed, a systematic method of producing the desired quality of coverage.

The other value of this approach is that by simply maintaining the stand polygon layer, getting a new Landsat scene and maintaining a minimal stand exam program this high quality current vegetation layer can be maintained quickly and inexpensively from year to year.