The optimum seeding rate of spring wheat (Triticum aestivum) can vary depending on the cultivar grown and the environment. Research to determine the optimum seeding rate of new varieties can be expensive. The objective of this research was to determine if the optimum seeding rate of a new cultivar in a given environment could be predicted using genetic and phenotypic traits and characteristics of the environment. In 32 different environments in North Dakota and Minnesota, a diverse selection of 21 hard red spring wheat cultivars differing in phenotype (e.g. plant height, tillering capacity, days from planting to heading) and genotype (Rht-B, Rht-D, Ppd-D1) were grown at four or five seeding rates. The optimum seeding rates were determined for each cultivar in each environment, from linear, quadratic or quadratic plateau equations. Recursive partitioning decision trees were fit with the ‘caret’ package using R Statistical Software using the optimum seed rate as the dependent variable and genotypic and phenotypic characteristics of the varieties and the yield potential and location of the environments as independent variables. The primary branch of the final model was dependent on location (longitude). Secondary branches included average yield (of growing environment) and varietal straw strength. Additional branch points occurred based on varietal tillering characteristics, Rht-B gene, latitude, days from planting to heading, and planting day of the year. Optimum predicted seeding rates ranged from 2.5 to 4.6 million seeds ha-1. The root of the decision tree (i.e. irrespective of cultivar, location or timing of seeding) indicated optimum seeding rate of 4.0 million seeds ha-1. This decision tree shows promise in recommending seeding rates for specific varieties in specific environments that are more productive and input efficient, than the “average” rate currently recommended. 018985