Fig. 8: Comparison of multi-scale training performance under different input orders.

This figure compares the prediction performance of two multi-scale training schemes—large-to-small (first row) and small-to-large (second row)—at both macro and micro landscape scales.