*3.1. Modeled Ya*

Pixel-level Ya in 2010–2015 over the NCP was simulated using the modified PRYM– Maize. Results are shown in Figure 4. In each year, modeled prefecture-level yields had a root mean standard deviation (RMSD) value of 0.88 ± 0.14 t hm−<sup>2</sup> and an R value of 0.49 ± 0.24 with the reference value. A pooled analysis for modeled yield vs. reference yield in 2010–2015 shows modeled yield by PRYM–Maize has an R (RMSD) of 0.45 (0.87 t hm−2). The performance of PRYM–Maize is improved when modeled and reference prefecture-level yields are averaged in time (Figure 5b). Uncertainties in modeled yield are significantly reduced, and the RMSD of mean annual modeled yield vs. the reference value is 0.66 t hm−2, which is less than that in any one year (Figure 4).

The average of R (RMSD) values for modeled yield vs. reference value in each year is 0.49 (0.86 t hm−2). The model achieves the best simulations in 2011 with an R value of 0.66. Generally, R values achieved by PRYM–Maize in this study are not high, and one important reason for this is that the prefecture-level maize yield values in 2010–2015 are in a narrow range (4–9 t hm−2) and show relatively small dynamics over space, as represented by both reference (standard deviation (STD) = 0.88 t hm−<sup>2</sup> for reference yield) and modeled yield (Figures 4 and 5a). Therefore, these results demonstrate that PRYM–Maize outputs reasonable estimates for the Ya of summer maize over the NCP.
