2.2.2. Meteorological Data

Gridded daily meteorological data retrieved from the ERA-Interim reanalysis (ERA) dataset and multi-site data of the China Meteorological Administration (CMA) were required to drive the RS crop yield model. Surface net radiation (Rn), vapor pressure

deficit (VPD), and air temperature (T) were directly retrieved from the ERA-Interim dataset (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era-interim, accessed in May and June, 2020). Precipitation (Pr) and global solar radiation (Rg) were retrieved from CMA multi-site data (http://data.cma.cn/, accessed in June 2020). Site-scale CMA data were spatially interpolated to a raster dataset using inverse distance weighted method provided by ArcGIS software (v10.1), and key parameters required were set as follows:



Rather than directly retrieving from the site-scale observations, daily solar radiation (*R*g) data were calculated in terms of the interpolated site-observed daily air temperature range (*T*R) and daily sun hours (*Hr*S) because sites observing *R*g is too sparse to be interpolated. We referred to Chen, et al. [39] to calculate daily *R*g, such that

$$R\text{g/}R\_0 = a \times \ln(T\_\text{R}) + b \times \left(Hr\_\text{S}/Hr\_\text{day}\right)^c + d \tag{2}$$

where *R*0 denotes the extra-terrestrial solar radiation; *Hr*day denotes the number of daytime hours; and *a*, *b*, *c*, and *d* are empirical coefficients. The average of values for each coefficient across multiple sites over China was used, such that *a* = 0.04, *b* = 0.48, *c* = 0.83, and *d* = 0.11. ERA-Interim datasets provide gridded global meteorological variables from 1981 to present in multiple temporal and spatial resolutions. The temporal and spatial resolutions of gridded data retrieved from ERA-Interim were 12 h and 0.125 arc-degree. The daily value of each variable is the sum (for Pr) or average (for variables except for Pr) of two 12-h values in one day.

#### 2.2.3. Reference Maize Yield

Prefecture-level statistics of maize yield in the period 2010 to 2015 reported by the National Bureau of Statistics of China were used to validate Ya simulated by a processbased and RS-driven crop yield model for maize (PRYM–Maize). These data were retrieved from the statistical yearbook of Shandong Province, Hebei Province, Henan Province, Beijing, and Tianjin for 2010–2015. When validating the simulations against statistical values, simulated pixel-level yields were averaged by prefecture-level districts (Figure 1).
