home > data and maps > supplemental data > forest stand

ForestERA Data Layer Overview - Forest Service Stand Polygons

Description

This is a vector dataset describing basal area (ft2/acre), tree stem density (trees/acre), canopy cover (%), and dominant overstory vegetation in USDA Forest Service stand polygons, across the western Mogollon Plateau in Arizona.

Purpose

The data contained in this layer were created as part of the Forest Ecosystem Restoration Analysis (ForestERA) project to support landscape-scale forest restoration planning efforts by a broad group of stakeholders including federal and state agencies, academic institutions, and non-governmental entities. Specifically, this layer was designed to meet the needs of the USDA Forest Service and other federal agencies that use forest service stand polygons as management and analysis units. These data are intended for regional analyses over spatial extents on the order of tens to hundreds of thousands of acres, and were not developed for use at finer spatial scales, although they may be useful for some applications at finer scales.

Development

This dataset was produced using zonal statistics in ArcGIS spatial analyst using USDA Forest Service stand polygons as zones to summarize 90m raster datasets created by the Forest Ecosystem Restoration Analysis (ForestERA) project. These raster datasets, representing basal area (m2/ha), stand density (trees / ha), canopy cover (%), and dominant overstory vegetation were created by the ForestERA project for the western Mogollon Plateau. The original vector dataset containing the forest stand boundaries was created by merging USDA Forest Service vector datasets in ArcGIS. Stand boundaries from the Coconino, Kaibab, Tonto, and Apache-Sitgreaves National Forests, that fell within the ForestERA western Mogollon Plateau assessment area were combined into a single vector dataset. Additional polygons were added which cover areas, such as parcels of private land, that fall within the western Mogollon Plateau study area but were not included in the Forest Service datasets.

The original ForestERA raster datasets were converted from their UTM NAD83 projection to UTM NAD27 projection to match the data typically used by the USDA Forest Service and other government agencies. In addition, the basal area and tree stem density datasets were converted from metric to English units to match the data used by government agencies. The raster datasets were then overlaid with the vector dataset containing the forest stand boundaries. The ArcGIS zonal statistics function was used to determine attributes of the raster datasets within the boundaries of each forest service stand. These attributes were then added to the database table for the vector dataset.

The attributes included for basal area (BA), tree stem density (TD), and canopy cover (CC), were mean (MEAN), maximum (MAX), minimum (MIN), and standard deviation (STD), of the values for the 90m cells that overlapped each stand polygon (capitalized codes in parentheses here and after correspond to the codes in the data table for the stand polygons). For example, if a forest service stand polygon overlapped ten 90m cells, then the value for mean canopy cover in that forest service stand was determined by obtaining the mean value for those 10 cells.

For dominant overstory vegetation we included a different set of attributes. The raster layer of dominant overstory vegetation included 9 vegetation classes; (1) open grasslands, shrublands, and barren areas (OPEN), (2) ponderosa pine (Pinus ponderosa; PIPO), (3) quaking aspen (Populus tremuloides, ASPEN), (4) mixed-conifer (MIXEDCON), including mixtures of spruces, firs, pines and/or Douglas’ Fir (Pseudotsuga menziesii), (5) pinyon-juniper (Pinus edulis and Juniperus spp.; PJ), (6) juniper-dominated mix (JUNIPINE), (7) ponderosa pine / quaking aspen (PINEASP), (8) ponderosa pine / Gambel’s oak (Quercus gambelii; PINEOAK), and (9) mixed-conifer / quaking aspen (MCASPEN). For the vegetation type field in the vector dataset we used the vegetation class with the highest number of cells that overlapped the stand polygon (DOMVEG). However, we also included the percentage of cells of each vegetation type within the stand. Consider as an example a stand that overlaps 10 cells, of which 6 are identified as ponderosa pine and 4 are identified as aspen. In this case we called ponderosa pine the dominant vegetation in the stand. However we also included attributes which show that 60% of the cells in the stand were identified as ponderosa pine and 40% were identified as aspen.

Accuracy Assessment

We did not undertake an accuracy assessment of this layer. Details on the accuracy and uncertainty of the original raster datasets may be obtained from the metadata files included with those layers. We provide here a brief description of the accuracy and uncertainty of each of the layers.

Canopy Cover

Accuracy assessment (linear regression) indicated that the relationships between the estimated values of canopy cover in the ForestERA raster dataset and actual values for canopy cover on the ground were highly significant and nearly unbiased (n = 200, r² = 0.545, P < 0.0001, m = 1.02). Based on uncertainty analysis we have estimated that values for 50% of the pixels in the DOQ derived canopy cover layer will be within 9% of the actual values for canopy cover in those areas, and values for over 80% of the pixels will be within 16% of the actual values.

Basal Area

Accuracy assessment (linear regression) indicated that the relationships between the estimated values of basal area in the ForestERA raster dataset and actual values for basal area on the ground were highly significant and nearly unbiased (n = 63, r² = 0.508, P < 0.0001, m = 1.09). Based on uncertainty analysis we have estimated that values for 50% of the pixels in the DOQ derived basal area layer will be within 4 m2/ha (17.5 ft2/acre) of the actual values for basal area in those areas, and values for over 80% of the pixels will be within 7m2/ha (30.5 ft2/acre) of the actual values.

Tree Stem Density

Accuracy assessment (linear regression) indicated that the relationships between the estimated values of tree stem density in the ForestERA raster dataset and actual values for tree stem density on the ground were highly significant and nearly unbiased (n = 63, r² = 0.584, P < 0.0001, m = 0.99). Based on uncertainty analysis we have estimated that values for 50% of the pixels in the DOQ derived tree stem density layer will be within 80 trees/ha (33 trees/acre) of the actual values for tree stem density in those areas, and values for over 80% of the pixels will be within 150 trees/ha (60 trees/acre) of the actual values.

Dominant Overstory Vegetation

We provide two metrics for accuracy assessment, the misclassification error rate, and the Kappa value. It has been suggested that an overall internal misclassification error rate of no more than 15%, and an individual category misclassification error rate of no more than 30%, should be adopted by researchers as a standard for all land classification efforts (Thomlinson et al. 1999). Our vegetation layer met these standards. Kappa values are a measure of agreement between datasets that takes into account the expected rate of agreement between those datasets based on chance alone (Foody 2002). Kappa values over 0.75 indicate excellent correspondence between the predicted and actual vegetation types while values of above 0.5 indicate good correspondence (Mainji et al. 2000). Based on these thresholds, the internal accuracy assessment indicated that the results of the modeling effort were excellent (Kappa = 0.91), while cross-validation results (Kappa = 0.57), and external accuracy assessment (Kappa = 0.54) which are generally considered more representative of actual accuracy, indicated the results of the modeling effort were good.

Sources of Errors

In general, the accuracy of the estimates for canopy cover, basal area, and tree stem density within the stand boundaries in this layer should be somewhat better than those of the original raster dataset. This is because errors (overestimation and underestimation) in individual cells should cancel each other out to some degree. However, any data obtained from an overlay process is likely to gain some error due to imprecision of the overlay. Errors may be greater overall for dominant overstory vegetation since it is difficult to categorize stands as a single vegetation type when the raster cells are themselves categorizations. Depending on the application, this may need to be taken into consideration when classifying the vegetation in a stand.

Recommendations

We believe this data layer can be used for the same purposes and applications as current forest service stand exam data.

Literature Cited

Foody, G. M. 2002. Status of land cover classification accuracy assessment. Remote-sensing of the environment 80: 185-201.

Maingi, K. John, S. E. Marsh, W. G. Kepner, and C. Edmomds. 2000. An accuracy assessment of 1992 landsat-MSS derived land cover for the upper San Pedro watershed (U.S/Mexico). EPA Research Report (EPA/600/R-02/040), United States Environmental Protection Agency. 21pp.

Thomlinson, J. R., P. V. Bolstad, and W. B. Cohen. 1999. Coordinating methodologies for scaling landcover classification from site-specific to global: steps toward validating global map products. Remote Sensing of the Environment 70: 16-28.

Please contact the ForestERA team to have a CD-ROM of this data mailed to you.

Page last updated February 11, 2005