Short-Term Time Step Convergence in a Climate Model


Due to constraints on computing resources, weather and climate calculations can only be done at finite—and often coarse—temporal resolutions, inevitably causing error. A novel technique, developed by scientists at Pacific Northwest National Laboratory, Sandia National Laboratories, and University of Michigan, efficiently quantified and attributed time-resolution errors in the Community Atmosphere Model version 5 (CAM5). Their work is the first publication to evaluate the time-step convergence, namely the reduction of numerical error as a result of a decrease in time-step length, in its strict mathematical sense, in a full-fledged atmospheric general circulation model. This is also the first attempt in the climate modeling community to quantitatively compare time-stepping errors, associated with different physical processes, in a model’s operational configuration. The team found that the temperature error in CAM5 converges at a rate of 0.4 instead of 1.0, indicating the error does not decrease as quickly as expected when the temporal resolution is increased. They performed sensitivity simulations to evaluate various subgrid-scale physical parameterizations in isolation. These simulations led to the conclusion that the representation of stratiform clouds is the primary source of time-stepping error in CAM5. The research showed that in this model, processes associated with the slowest convergence rates also produced the largest errors and strongest artificial sensitivities. Slow convergence is thus a ‘‘?ag’’ for model components that do not accurately represent the intended physical balance of processes and require more attention for improvement.


Wan, H., P. J. Rasch, M. A. Taylor, and C. Jablonowski. 2015. “Short-Term Time Step Convergence in a Climate Model,” Journal of Advances in Modeling Earth Systems 7(1), 215–25. DOI: 10.1002/2014MS000368.