Assessment of climate biases in OpenIFS version 43r3 across model horizontal resolutions and time steps

Abstract

Abstract. We examine the impact of horizontal resolution and model time step on the climate of the OpenIFS version 43r3 atmospheric general circulation model. A series of simulations for the period 1979–2019 are conducted with various horizontal resolutions (i.e. ∼100, ∼50, and ∼25 km) while maintaining the same time step (i.e. 15 min) and using different time steps (i.e. 60, 30, and 15 min) at 100 km horizontal resolution. We find that the surface zonal wind bias is significantly reduced over certain regions such as the Southern Ocean and the Northern Hemisphere mid-latitudes and in tropical and subtropical regions at a high horizontal resolution (i.e. ∼25 km). Similar improvement is evident too when using a coarse-resolution model (∼100 km) with a smaller time step (i.e. 30 and 15 min). We also find improvements in Rossby wave amplitude and phase speed, as well as in weather regime patterns, when a smaller time step or higher horizontal resolution is used. The improvement in the wind bias when using the shorter time step is mostly due to an increase in shallow and mid-level convection that enhances vertical mixing in the lower troposphere. The enhanced mixing allows frictional effects to influence a deeper layer and reduces wind and wind speed throughout the troposphere. However, precipitation biases generally increase with higher horizontal resolutions or smaller time steps, whereas the surface air temperature bias exhibits a small improvement over North America and the eastern Eurasian continent. We argue that the bias improvement in the highest-horizontal-resolution (i.e. ∼25 km) configuration benefits from a combination of both the enhanced horizontal resolution and the shorter time step. In summary, we demonstrate that, by reducing the time step in the coarse-resolution (∼100 km) OpenIFS model, one can alleviate some climate biases at a lower cost than by increasing the horizontal resolution.

Impact of horizontal resolution and model time step on European precipitation extremes in the OpenIFS 43r3 atmosphere model

“…6a). However, the sensitivity is not found for the global-mean total precipitation by increasing horizontal resolution (Savita et al, 2024). For the large-scale precipitation in JJA, the CI for RMSEs from LR do not overlap with those from MR and HR at higher percentile ranges (>95 th ), but overlap at lower percentile ranges (<95 th ) (Fig.…”

“…We use the same AMIP simulations that were used in Savita et al (2024) which cover the period 1979-2014 and are extended to 2019 using sea-surface temperature (SST) from ERA5 and the Shared Socioeconomic Pathway 5 (SSP5-8.5) scenario from CMIP6. OpenIFS simulations use 91 vertical levels (L91) and the different horizontal resolutions: low resolution (Tco95, ~100 km), medium resolution (Tco199, ~50 km), and high resolution (Tco399, ~25 km).…”

“…The LR, LR30m and LR60m data were interpolated to a global 0.9° grid while the MR and HR data were interpolated to a global 0.45° grid, i.e., we are not investigating extreme precipitation in high resolution simulations in their native grid, which will be investigated in future study. The simulations used here were used by Savita et al (2024) who found improvements in the surface zonal wind, Rossby wave amplitude and phase speed, weather regime patterns, and surface-air temperature when reducing a model time step from 60 minutes to 30 and 15 minutes in low resolution or increasing the horizontal resolution from 100 km to 50 and 25 km. However, Savita et al (2024) did not find such improvement in the mean precipitation bias by increasing horizontal resolution or reducing the model time step.…”

“…The simulations used here were used by Savita et al (2024) who found improvements in the surface zonal wind, Rossby wave amplitude and phase speed, weather regime patterns, and surface-air temperature when reducing a model time step from 60 minutes to 30 and 15 minutes in low resolution or increasing the horizontal resolution from 100 km to 50 and 25 km. However, Savita et al (2024) did not find such improvement in the mean precipitation bias by increasing horizontal resolution or reducing the model time step.…”

“…A recent study by Savita et al (2024) explored the sensitivity of global mean precipitation to the horizontal resolution and model time step in atmosphere-only simulations with OpenIFS.…”