ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
Archive > Volume 46 Issue 3 > 2020,46(3):367-380. DOI:10.7519/j.issn.1000-0526.2020.03.008 Prev Next
Comprehensive Evaluations of GRAPES_3 km Numerical Model in Forecasting Convective Storms Using Various Verification Methods
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
GRAPES_3 km (Global/ Regional Assimilation and Prediction System） is a convection-permitting model, which provides an important objective basis for short-term forecast of severe weather. In this paper, the performance of GRAPES_3 km model in severe weather forecasting was comprehensively evaluated using traditional pixel-versus-pixel threat score, neighborhood and object-based methods. The applicability and differences of the traditional and new spatial verification methods for high resolution model assessment were analyzed, and the results were compared with those of GRAPES_Meso model. The results are showed as follows. Through case analysis, it is found that the characteristics of convective storm forecast and the evolution of convection 〖JP2〗can be comprehensively and objectively evaluated by various verification methods. GRAPES_ 3 km is superior to GRAPES_Meso in forecasting convective storms, especially severe storms over 50 dBz. The latest forecast of the initial forecast time is the best. Neighborhood TS method takes space-time deviation into account.The forecasting skill is the highest when the time neighborhood of GRAPES_3 km model is 1 h for 20 dBz and 35 dBz, and 3 h for 50 dBz. The fractions skill score (FSS) shows that GRAPES_3 km model can achieve the lowest forecast skill scale for convective storms with different thresholds, while GRAPES_Meso model usually fails to reach the lowest skill scale for storms above 35 dBz. Method for object-based diagnostic evaluation (MODE) can be used to evaluate the forecast of convective storm attributes. GRAPES_3 km model is consistent with the actual number of storms at all scales, but the area is obviously underestimated. The model can predict the shape and location of meso-β scale convective storms, while for meso-γ scale convective storms, the forecast scale is larger, the shape is more circular and the axis angle is smaller, but conversely for meso-α scale convective storms. The traditional pixel-versus-pixel verification method and new spatial verification methods have the same conclusion for convection-permitting model. Comparatively, the new spatial verification methods can provide more detailed information of convective storms.
Keywords:
Project Supported:
Clc Number:
Mobile website
