Abstract |
To investigate the predictability for a sudden severe rainstorm event occurred on September 11, 2014 around Kobe city, we perform a series assimilation (DA) experiments using the Local Ensemble Transform Kalman Filter (LETKF) with the JMA-NHM (NHM-LETKF). In the experiments, we used Phased array weather radar (PAWR) at Osaka University and surface data observed by the original weather stations “POTEKA II” at 8 locations in Kobe city every 30 seconds. Since the surface station data had significant biases, a bias correction method has been developed with the Kobe observatory data as the unbiased ground truth.
Three DA experiments with the PAWR (NO-POTEKA), with the PAWR and the raw data of POTEKA II (NOBC), and the PAWR and the bias corrected data (BC) were performed at 1-km. In NO-POTEKA, strong echoes and surface rainfalls were seen, although the rainfall intensity is smaller than the JMA analyzed precipitation based on the radar and gauge networks. NOBC showed a significant decrease in surface relative humidity because of the dry biases of the surface station data. By contrast, BC showed stronger rainfall intensity, better matching with the JMA analyzed precipitation.
In addition, we also performed a high resolution experiments at 100-m have been performed with the PAWR. In the result, detailed structure of active convections were clearly reproduced and surface rainfall amount was significantly improved compared to the experiments at 1-km resolution.
From the series of the DA experiments, we suggest that the dense and frequent surface DA have a potential to improve the forecast accuracy for sudden severe rainstorms. And the horizontal resolution at 100-m also plays an important part in the numerical weather forecast for a sudden severe rainstorm.
|