Potential natural vegetation classification and mapping for Washington, Oregon, and California using Gradient Nearest Neighbor imputation

Abstract

Potential natural vegetation (PNV) classification and mapping are foundational tools in planning in forestry and related fields to identify appropriate biophysical environments for vegetation communities across the landscape. The need for this information has grown over recent decades as attention has shifted to planning over large geographic areas. Our goal of this work is to develop a consistent, continuous potential vegetation map of both forest and nonforest for Washington, Oregon, and California.

We built a PNV map using wall-to-wall, 30-m maps of current forest attributes from the Gradient Nearest Neighbor (GNN) imputation vegetation dataset developed by the Landscape Ecology, Modeling, Mapping, and Analysis (LEMMA) group, a collaborative research partnership between the U.S. Department of Agriculture, Forest Service and Oregon State University. The vegetation zone (vegzone) and subvegetation zone (subzone) layers were derived from overstory and understory species composition and abundance (percentage of cover) in the GNN maps. We took advantage of the entire temporal span of GNN (1986–2017) to maximize chances of successfully inferring potential vegetation from current vegetation.

For forest vegzone assignments, we used presence and percentage of cover of tree species. We established a hierarchical ranking of the tree species identified as indicator species within plant association guides and vegetation manuals for the region. The ranking of tree species is based on shade tolerance, ecological amplitude, and longevity. To assign subzones to inventory plots that have already been classified to a vegzone, we considered graminoid, herb, shrub, and tree species that were not used in vegzone assignments. Using both the Forest Service, Forest Inventory and Analysis (FIA) plots and the Pacific Southwest Region and Pacific Northwest Region ecology plots, we examined species distributions with respect to climatic variables and their similarities to one another in species-composition space. Based on this analysis, we established nine species groups indicating different combinations of temperature and moisture regimes and a separate list of species that indicated serpentine substrates.

We developed a standardized classification scheme to create and map potential vegetation units across large, multistate landscapes. Results of an accuracy assessment indicate that the maps also have utility at finer scales (thousand to hundreds of thousands of acres), with appropriate consideration of spatial variation in map accuracy. This PNV product is an appropriate framework for additional, complementary, spatial products that can be used to study existing vegetation, disturbance regimes, and successional pathways.

Supplemental Download:

Appendix 6: Species Groups and Key Indicator Species (Bin Species)

Appendix 9: Vegzone and Subzone Abundance and Distribution in Level III Ecoregions

Appendix 13: Vegzone Error Matrix

Appendix 14: Subzone Error Matrix

Appendix 15: Vegzone Error Matrices by Gradient Nearest Neighbor Modeling Region

Appendix 16: Subzone Error Matrices by Gradient Nearest Neighbor Modeling Region

Appendix 17: Area-Based Regional Accuracy Assessment

Citation

Simpson, Michael; Kertis, Jane; Acker, Steven A.; Hochhalter, Patricia A.; Bell, David M.; Gregory, Matthew J.; DeMeo, Thomas. 2026. Potential natural vegetation classification and mapping for Washington, Oregon, and California using Gradient Nearest Neighbor imputation. Gen. Tech. Rep. PNW-GTR-1041. Portland, OR: U.S. Department of Agriculture, Forest Service, Pacific Northwest Research Station. 295 p. https://doi.org/10.2737/pnw-gtr-1041.