Historical reference conditions have provided empirical benchmarks for guiding ecological restoration but the relevance of historical conditions has been questioned in this era of rapid global change. Trait-based ecology offers an alternative approach for setting restoration objectives. If we understand which phenotypic traits confer high survival in a given environment, then we can restore assemblages of species with favorable trait combinations to reduce mortality risk, achieve functional outcomes, and enhance restoration success. Our objective was to compare restoration prescriptions based on historical reference conditions versus trait-based objectives in southwestern United States mixed conifer forests. To optimize survival and enhance resilience under projected increasing frequencies of fire and drought, we constructed and evaluated models based on combinations of three traits: thick bark, dense wood, and moderate leaf nitrogen concentration. Models with multiple traits accurately derived historical species abundance distributions, which is a necessary condition for the application of trait-based models under less certain future conditions. Model results indicated that trait-based restoration objectives could be achieved in two ways: by manipulating abundances of species that already coexist at a site or by adding native species from warmer climates to the local species pool. The latter approach may create no-analog communities of native species. Restoration goals based on either historical reference conditions or future projections are special cases of a more general class of desired future conditions that are derived from trait-based objectives. Functional ecology provides a general, flexible, and theory-based approach to restoring resilient ecosystems at a time of rapid environmental change.