Least-squared linear regression results of the relationships between radar estimates of growing season parameters and model simulations of annual NPP, GPP and Raut are summarized in Table 6. Radar estimates of the initiation of the 2000 and 2001 growing seasons showed significant ( p < 0.001), inverse relationships with BIOME-BGC annual results, accounting for 78.5%, 80.9% and 57.0% of spatiotemporal variations in estimated GPP, Raut and NPP, respectively. Radar estimates of growing season cessation were not significantly associated with model results, while correspondences between radar estimates of annual growing season length and model results were significant, but generally lower than estimates of growing season initiation. Associations between radar remote sensing results and NPP were also somewhat weaker than relationships with GPP and Raut, primarily because NPP is derived as a residual of the other two much larger terms and because of the similar inverse sign and slope of the relationships between these variables and radar results. Correspondences between radar estimates of growing season length and BIOME-BGC results were also comparatively weak because of the apparent importance of the timing of spring thaw in regulating annual productivity relative to the cessation of the growing season.