干旱气象 ›› 2023, Vol. 41 ›› Issue (2): 328-340.DOI: 10.11755/j.issn.1006-7639(2023)-02-0328

• 技术报告 • 上一篇    下一篇

GRAPES-REPS对我国南方2017年初夏持续性降水预报的检验评估

王叶红(), 赵玉春   

  1. 福建省厦门市气象局,福建省厦门市海峡气象开放重点实验室,福建 厦门 361012
  • 收稿日期:2022-03-24 修回日期:2022-08-12 出版日期:2023-04-30 发布日期:2023-05-09
  • 作者简介:王叶红(1974—),女,河南许昌人,正高级工程师,主要从事资料同化和中尺度数值模拟研究。E-mail: 278403168@qq.com
  • 基金资助:
    福建省自然科学基金项目(2021J01466)

Verification and assessment of persistent rainfall forecasts of GRAPES-REPS in pre-summer of 2017 in southern China

WANG Yehong(), ZHAO Yuchun   

  1. Xiamen Meteorological Bureau of Fujian Province, Xiamen Key Laboratory of Straits Meteorology, Xiamen 361012, Fujian, China
  • Received:2022-03-24 Revised:2022-08-12 Online:2023-04-30 Published:2023-05-09

摘要:

我国自主研发的GRAPES-REPS(Global/Regional Assimilation and Prediction System-Regional Ensemble Prediction System)于2014年投入业务运行,为加深对该系统降水集合预报能力的认识,便于更好应用降水概率预报,本文以2017年5月中旬至6月下旬我国南方地区3次持续性降水过程为例,采用统计检验和个例分析相结合,评估该系统在72 h内不同预报时效的各量级24 h降水预报性能。结果表明:(1)GRAPES-REPS集合平均预报对小雨、中雨有较明显优势,但随着降水量级增大优势逐渐下降,对暴雨预报已不具有优势。其中,小雨预报范围接近观测,中雨(暴雨)有空报(漏报)倾向,大雨在较长预报时效上有空报倾向。(2)采用MRF边界层参数化方案和KF-eta积云对流参数化方案组合方案的控制预报成员和2个扰动预报成员为集合最优成员,其TS评分普遍高于采用其他组合方案的集合成员。(3)整体上集合成员降水预报离散度不足,尤其0~24 h预报时效,Talagrand分布呈U型,对小(大)量级降水预报概率偏大(小);随着预报时效增加,集合成员预报离散度显著增大,Talagrand分布逐渐接近理想概率分布。(4)集合预报在不同时效对各量级降水概率预报均具有参考价值,大雨、暴雨的概率预报效果优于小雨、中雨。(5)集合预报整体上能够较好把握典型暴雨日降水空间分布形态,对中央气象台漏报的广东中南部暖区暴雨有一定的概率预报能力。

关键词: GRAPES-REPS, 集合预报与检验, 初夏降水, 我国南方

Abstract:

The self-developed global/regional assimilation and prediction system-regional ensemble prediction system (GRAPES-REPS) was put into operation in 2014 in China. In order to deeply understand the precipitation ensemble forecast ability of this system and conveniently apply the precipitation probability forecast, in this paper, the 24 h accumulated precipitation with different magnitudes forecasted by GRAPES-REPS at different lead time within 72 hours is evaluated by using statistical analysis and case analysis taking three continuous precipitation processes in southern China from mid-May to late June 2017 for example. The results are as follows: (1) The ensemble mean forecast of GRAPES-REPS has obvious advantage for light rain and moderate rain. The advantage decreases gradually with the increase of precipitation magnitude and no advantage appears for rainstorm. The ensemble mean forecast is close to the observation for light rain, while it has a tendency of null (missing) forecast for moderate rain (rainstorm) or heavy rain at a longer lead time. (2) The optimal members include control forecast and two perturbation forecasts that use a combination of MRF boundary layer scheme and KF-eta cumulus convection scheme, which is different to the other members. (3) The spread of precipitation ensemble forecasts is insufficient overall, especially at 0-24 h forecast lead time with the U-shaped Talagrand distribution and the higher (lower) forecast probability for small- (large-) magnitude precipitation. The spread increases obviously with the increase of forecast lead time, and the Talagrand distribution is gradually close to the expected-probability distribution. (4) The ensemble forecasts do have a reference value for precipitation with different magnitudes at every forecast lead time, with the probability forecast of heavy rain and rainstorm being better than that of light rain and moderate rain. (5) The ensemble gives a better forecast for precipitation pattern with different magnitudes as a whole, especially it has an ability of probability forecast for the warm-sector rainstorms in the central and southern Guangdong Province, which is missing in the forecasts of National Meteorological Observatory.

Key words: GRAPES-REPS, ensemble forecast and verification, pre-summer precipitation, southern China

中图分类号: