一次高温事件的GRAPES对流尺度模式预报能力评估

Evaluation of the GRAPES Convective Scale Model Prediction Capability for a High Temperature Event

摘要: 为了评估GRAPES对流尺度模式对高温事件的预报能力，本文选取成都市及其周边区域(102˚E~105˚E, 30˚N~31.5˚N)，7月28日至8月4日的GRAPES 3 km模式温度预报资料、ERA5温度再分析资料和成都市各测站常规资料中的温度资料，利用图像识别的特征区域提取算法提取出温度高于30℃的区域，通过计算相关系数等对此次成都市的高温事件进行GRAPES对流尺度模式高温事件预报能力评估。本文从四个方面进行讨论：高温事件的开始时间、结束时间、高温区域面积和高温强度。结果表明：1) GRAPES对流尺度模式预报高温区域中心变化趋势与ERA5再分析资料的高温区域中心变化趋势相一致，但GRAPES对流尺度模式对高温区域中心的预报与ERA5再分析资料相比整体偏北、偏西。2) GRAPES对流尺度模式对高温出现和结束的时段的预报与ERA5再分析资料高温出现和结束时间相比均有所提前，高温开始时间提前1~2小时，高温结束时间提前1~3小时。3) GRAPES对流尺度模式预报高温区域面积与ERA5再分析资料相比整体偏大，二者高温面积相关性较高，高温区域面积的预报能力较好。4) GRAPES对流尺度模式出现高温时段温度变化趋势与ERA5再分析资料相符合，GRAPES对流尺度模式对高温强度的预报与ERA5再分析资料的高温强度相比整体较弱。

Abstract: This paper selects the temperature data from the GRAPES 3 km model prediction of temperature data, the ERA5 reanalysis temperature data and the conventional data from Chengdu stations from July 28 to August 4 in Chengdu and its surrounding areas (102˚E~105˚E, 30˚N~31.5˚N), to evaluate the prediction ability of the GRAPES convective scale model for high temperature events. By using the feature area extraction algorithm of image recognition to extract the area with temperature higher than 30˚C, calculating the correlation coefficient to evaluate the capability of forecasting the high temperature event of the GRAPES convective scale model for this high temperature event in Chengdu. This paper discusses from four aspects: the start time, the end time, the area of high temperature region and the intensity of high temperature region of this high temperature event. Major results are as follows. 1) The prediction trend of the regional center of high temperature by the GRAPES convective scale model is consistent with that by ERA5 reanalysis data, but the prediction trend of the regional center of high temperature by the GRAPES convective scale model is generally more north and more west than that by ERA5 reanalysis data. 2) Compared with ERA5 reanalysis data, the prediction of the onset and end of high temperature by the GRAPES convective scale is earlier. The GRAPES convective scale model predicts that the onset of high temperature is 1~2 hours earlier and the end of high temperature is 1~3 hours earlier. 3) The GRAPES convective scale model has a high correlation with ERA5 reanalysis data, the prediction ability of high temperature area is better. 4) Compared with ERA5 reanalysis data, the GRAPES convective scale model predicted high temperature area is mainly larger than ERA5 reanalysis data. The correlation between the two models is high and the prediction ability of high temperature area is good.