P-36

Convective-scale predictability in numerical weather prediction at a 100-m resolution

S. Otsuka and T. Miyoshi (RIKEN)

Recent developments in high-performance computing and advanced observing technologies enable us to step forward to cumulus-convective-scale data assimilation for numerical weather prediction at a horizontal resolution of O(100) m, O(10) times higher than the previous studies (e.g., Leoncini et al. 2010; Melhauser and Zhang 2012; Keil et al. 2014). Understanding the predictability of convective-scale weather plays an essential role in designing such high-resolution numerical weather prediction systems. In particular, it would be important to know what would be the effective temporal frequency of data assimilation, whether or not it needs to be the order of seconds. This study performs 30-second breeding cycles at a 100-m resolution using the Weather Research and Forecasting (WRF) model, and explores the convective-scale predictability.

The results of breeding experiments show that bred vectors develop around the edge of newly developing convective cores and spread away from the convective cores. Different rescaling intervals with the same rescaling value show similar spatial structures of the bred vectors. By contrast, different rescaling values produce different spatial structures.