Figure 4. Distribution of mean absolute error (MAE) among phenoclimate models of a) population-level flowering onset DOY, b) population-level peak flowering DOY, and c) population-level flowering termination DOY trained using simulated species collected with a bias towards early individuals, collected without bias, or with bias towards late-flowering individuals within each local population. All species included in these models exhibited a phenological responsiveness of 4 Days/˚C, a high degree (σ = 30 days) of intrapopulation variation, and moderate (30 day) individual flowering durations. Within each panel, groups of models associated with different letters exhibit statistically different mean MAEs among groups of taxa. Where statistically significant differences in MAE were detected, statistical significance was high (p < 0.001) in all cases.