Name: SCF in Areas Managed for Timber Value (AMTV)
Display Field: AMEV
Type: Feature Layer
Geometry Type: esriGeometryPolygon
Description: DNR developed a detector model to classify westside DNR lands for the Landscape Assessment as structurally complex forest (SCF) or not. The requirements of the project were to: - use DNR's forest inventory (RS-FRIS) data and Stand Development Stage (SDS) field observations; - produce detections at RS-FRIS inventory scale (0.1 acre); - accommodate efficient re-fitting with new data, as needed; and - be compatible with growth and yield modeling software. Structurally Complex Forest consists of the Van Pelt stages: Maturation II, Vertical Diversification, Horizontal Diversification, and Pioneer Cohort Loss. DNR models for detecting SCF use combinations of the variables quadratic mean diameter of trees >= 6-in (qmd6) and canopy layers within 2.5-acre plot polygons. These models are applied to RS-FRIS rasters for DNR managed lands in Western WA to classify the land base. The models are applied by convolving a 2.5-acre circular kernel across forested areas within each RIU that contains data. The convolution implemented takes the average or standard deviation of all the data within a 2.5-acre circle around the center pixel. The outcome is that the each pixel is influenced by neighboring pixels. Model coefficients are then applied to the results of the convolution and the output is an image of probabilities of whether the pixel is SCF. The model accuracy using a 50% probability threshold for SCF vs. not SCF is 81% +/- 1%. These layers are DRAFT and subject to change. These layers should be used as a screening tool for structurally complex forest. They are not definitive. They are modeled detections where higher probabilities estimate a higher certainty of structurally complex forest. The models and GIS layers will change as the RS-FRIS inventory is updated and as the models are refit and retrained, as needed. References: Van Pelt, R. 2007. Identifying Mature and Old Forests in Western Washington. Washington Department of Natural Resources. 104 pp.
Description: DNR developed a detector model to classify westside DNR lands for the Landscape Assessment as structurally complex forest (SCF) or not. The requirements of the project were to: - use DNR's forest inventory (RS-FRIS) data and Stand Development Stage (SDS) field observations; - produce detections at RS-FRIS inventory scale (0.1 acre); - accommodate efficient re-fitting with new data, as needed; and - be compatible with growth and yield modeling software. Structurally Complex Forest consists of the Van Pelt stages: Maturation II, Vertical Diversification, Horizontal Diversification, and Pioneer Cohort Loss. DNR models for detecting SCF use combinations of the variables quadratic mean diameter of trees >= 6-in (qmd6) and canopy layers within 2.5-acre plot polygons. These models are applied to RS-FRIS rasters for DNR managed lands in Western WA to classify the land base. The models are applied by convolving a 2.5-acre circular kernel across forested areas within each RIU that contains data. The convolution implemented takes the average or standard deviation of all the data within a 2.5-acre circle around the center pixel. The outcome is that the each pixel is influenced by neighboring pixels. Model coefficients are then applied to the results of the convolution and the output is an image of probabilities of whether the pixel is SCF. The model accuracy using a 50% probability threshold for SCF vs. not SCF is 81% +/- 1%. These layers are DRAFT and subject to change. These layers should be used as a screening tool for structurally complex forest. They are not definitive. They are modeled detections where higher probabilities estimate a higher certainty of structurally complex forest. The models and GIS layers will change as the RS-FRIS inventory is updated and as the models are refit and retrained, as needed. References: Van Pelt, R. 2007. Identifying Mature and Old Forests in Western Washington. Washington Department of Natural Resources. 104 pp.