Abstract Storm‐resolving global climate models, approaching km‐scale grid spacings, are slowly emerging. With increasing horizontal resolution, we expect improvements in the surface radiation budget because (a) synoptic variability may be better represented, (b) convective systems are better resolved, and (c) landscape heterogeneity is represented in more detail. Here, we study the performance of two km‐scale global climate models within the next generation Earth Modeling Systems (nextGEMS) project, the Integrating Forecasting System (IFS) and the ICOsahedral Non‐hydrostatic model, in producing accurate downwelling surface solar and thermal irradiances. Furthermore, we study how horizontal resolution affects mean surface irradiances as well as irradiance variability on hourly time scales. Most simulations have global annual mean biases within 4.5 W and 7.6 W for solar and thermal irradiance, respectively. Locally, irradiance biases with respect to station observations are up to an order of magnitude higher, with no consistent improvement as we move to higher resolution. Comparing three IFS simulations ranging from 4.4 to 28 km resolution, we indeed find that differences in mean surface irradiance are most pronounced in areas with high orographic variability, suggesting that the better resolved orography is the main pathway of resolution‐driven changes in the surface irradiance climatology. Furthermore, we find that the temporal irradiance variability improves with horizontal resolution among these three simulations, but only at sub‐daily time scales. Our findings suggest that for an accurate representation of surface irradiance climatologies, moving towards km‐scale resolution may not improve much beyond the current generation of climate models.
Veerman et al. (2026) studied this question.