设为首页
加入收藏
The Optimal Training Period Scheme of MOS Temperature Forecast
-
摘要: 以欧洲中期天气预报中心 (ECMWF) 模式细网格地面气温为预报因子，设计多种训练期方案进行2014—2015年福建省气象站每日两次1~7 d的日最高气温和日最低气温MOS (model output statistics) 预报，并进行检验和改进。准对称混合滑动训练期方法为取预报日之前和前1年预报日之后相同日数的样本混合而成，分1年期或多年期。结果表明：准对称混合滑动训练期方案优于滑动训练期方案和传统季节固定期分类方案，且2年期优于1年期。以不同周期确定最佳训练期日数的方案应用对比显示，以年为评估周期优于以月为评估周期以及更短时间周期。在2015年日最高气温和日最低气温MOS预报中，基于上年度评估所得最佳训练期日数，2年期准对称混合滑动训练期方案较ECMWF模式细网格产品质量有较大提高，优于预报员预报，有较好的应用参考价值。
-
关键词:
- MOS预报;
- 日最高气温和日最低气温;
- 训练期;
- 准对称方法
Abstract: Based on air temperature from ECMWF data, three groups of training period schemes that include original schemes, improved schemes and application schemes are designed to test and compare results of daily maximum and minimum temperature of national stations in Fujian Province twice per day from 2014 to 2015. For the quasi-symmetrical mixed running training period (QSRTP) method, several days of current year before forecasting day and equal numbers of days in last year after are adopted separately to compose an initial 1-year or multi-year dataset. When using multiple-year data, lengths of the optimal training period every year are different or the same in each scheme. From three original schemes, it is found that the methods of QSRTP are much better than other running training period methods and traditional fixed period classification, and the number of optimal training day is more stable. The QSRTP with 2-year data shows better performance than that with 1-year data for comparison of improvement schemes, and different training period lengths lead to better performance in 2-year evaluation schemes. Considering model version updating and the continuity of weather, lengths of the optimal training period before the first year are slightly longer than that in the second year. Similar to the traditional method, the optimal training period of original and improved schemes are obtained after verification rather than before forecast. Three application schemes with different terms of evaluation are designed to test the stability, the usability and seasonal patterns of the optimal training period. Based on 1-year evaluation, total samples of the training period are stable with the best forecasting score. In terms of the monthly evaluation, the best period has no significant patterns with a relatively low score. For 10-day evaluation, the best period varies greatly, but when the forecasting time becomes shorter, the forecast quality becomes better. When the weather changes suddenly, the optimal training days will be quite different. Hindcasting experiment of 2015 suggests that the MOS forecast for daily maximum and minimum temperature using the optimal training period of last year has a much higher score than the original ECMWF products, it is better than the subjective forecast, and the forecast absolute deviation is significantly reduced. With the accumulation of data in the future, the forecast quality would be improved greatly, indicating that the method of multi-year QSRTP has an important application prospect on the daily operation. -
图 1 1~3 d初始方案2日最高气温 (a)、初始方案3日最高气温 (b)、初始方案2日最低气温 (c) 和初始方案3日最低气温 (d) 的预报准确率随滑动训练期变化
Fig. 1 Forecast accuracy in 1-3-day changes with running training days, daily maximum temperature of original scheme 2(a) and original scheme 3(b), daily minimum temperature of original scheme 2(c) and original scheme 3(d)
图 2 应用方案3的2015年4月日最高气温逐日最佳滑动训练期日数N
Fig. 2 The daily best running training days of maximum temperature forecast of application scheme 3 in Apr 2015
表 1 初始方案1~3的2014—2015年1~7 d日最高气温和日最低气温预报检验
Table 1 Verification results of daily maximum and minimum temperature 1-7-day forecast of original scheme 1, original scheme 2 and original scheme 3 from 2014 to 2015
要素 统计量 初始方案 1 d 2 d 3 d 4 d 5 d 6 d 7 d 日最高
气温均方根误差/℃ 1 1.96 2.12 2.36 2.53 2.69 2.89 3.13 2 1.92 2.11 2.38 2.58 2.76 2.96 3.22 3 1.90 2.09 2.35 2.55 2.72 2.93 3.19 预报准确率/% 1 74.91 71.81 67.31 63.37 60.41 57.41 53.86 2 77.16 73.39 68.24 63.84 60.42 57.11 52.74 3 77.93 74.29 68.98 64.46 61.05 57.54 53.44 日最低
气温均方根误差/℃ 1 1.33 1.42 1.59 1.74 1.83 1.92 2.03 2 1.30 1.40 1.56 1.71 1.82 1.91 2.04 3 1.27 1.36 1.51 1.65 1.75 1.83 1.97 预报准确率/% 1 89.38 87.73 84.01 80.55 78.6 76.71 73.84 2 89.82 87.74 84.25 81.27 78.99 76.65 73.53 3 90.62 88.97 85.84 82.55 80.30 78.00 74.81 
下载: 导出CSV
表 2 改进方案1~2的2014—2015年1~7 d日最高气温和日最低气温预报检验
Table 2 Verification results of daily maximum and minimum temperature 1-7-day forecast of improved scheme 1 and improved scheme 2 from 2014 to 2015
要素 统计量 改进方案 1 d 2 d 3 d 4 d 5 d 6 d 7 d 日最高
气温均方根误差/℃ 1 1.78 1.94 2.18 2.40 2.60 2.82 3.10 2 1.90 2.07 2.33 2.52 2.69 2.90 3.15 预报准确率/% 1 78.20 74.56 69.45 64.97 61.57 58.26 54.42 2 78.00 74.43 69.25 64.75 61.34 58.00 54.17 日最低
气温均方根误差/℃ 1 1.24 1.30 1.44 1.59 1.70 1.81 1.95 2 1.28 1.37 1.53 1.68 1.78 1.86 1.99 预报准确率/% 1 90.64 89.08 85.73 82.27 80.02 77.87 74.73 2 90.51 88.92 85.56 82.19 79.89 77.72 74.59
下载: 导出CSV
表 3 应用方案1~3的2015年1~7 d日最高气温和日最低气温预报质量检验结果
Table 3 Verification results of daily maximum and minimum temperature 1-7-day forecast of application scheme 1, application scheme 2 and application scheme 3 in 2015
要素 统计量 应用方案 1 d 2 d 3 d 4 d 5 d 6 d 7 d 日最高
气温均方根误差/℃ 1 1.81 1.97 2.24 2.47 2.69 2.96 3.31 2 1.82 1.98 2.25 2.47 2.69 2.96 3.31 3 1.81 1.97 2.25 2.47 2.7 2.97 3.33 预报准确率/% 1 78.07 74.23 68.73 63.81 59.86 56.85 52.12 2 78.06 74.17 68.57 63.78 59.71 56.5 51.97 3 78.14 74.29 68.71 63.79 59.81 56.71 52.04 日最低
气温均方根误差/℃ 1 1.23 1.31 1.44 1.55 1.69 1.85 2.03 2 1.23 1.3 1.44 1.55 1.69 1.85 2.03 3 1.23 1.3 1.44 1.56 1.7 1.86 2.04 预报准确率/% 1 91.19 89.6 86.35 83.74 80.86 78.04 73.90 2 91.22 89.65 86.45 83.88 80.80 78.01 73.82 3 91.49 89.67 86.20 83.40 80.63 77.58 73.51
下载: 导出CSV
表 4 2014年和2015年逐月最佳滑动训练期日数N(单位：d)
Table 4 The monthly best running training days (N) of daily maximum and minimum temperature forecast from 2014 to 2015(unit:d)
月份 日最高气温 日最低气温 2014年 2015年 2014年 2015年 1 31 36 44 21 2 42 22 21 21 3 29 23 44 23 4 22 26 21 35 5 38 21 39 45 6 23 20 23 21 7 22 32 29 45 8 22 21 30 26 9 40 23 45 24 10 21 45 24 45 11 22 32 38 43 12 34 23 41 23
下载: 导出CSV
表 5 ECMWF细网格产品、预报员、初始方案2和应用方案1在2015年1~7 d日最高气温和日最低气温订正预报检验
Table 5 Verification results of daily maximum and minimum temperature 1-7-day forecast produced by ECMWF, forecasters, original scheme 2 and application scheme 1 in 2015
要素 时效/d ECMWF细
网格产品预报平均绝对偏差/℃ 预报技巧评分/% 预报员 初始方案2 应用方案1 预报员 初始方案2 应用方案1 日最高
气温1 2.46 1.40 1.41 1.35 43.09 42.68 45.12 2 2.53 1.62 1.54 1.48 35.97 39.13 41.50 3 2.63 1.80 1.75 1.68 31.56 33.46 36.12 4 2.79 2.01 1.92 1.86 27.96 31.18 33.33 5 2.89 2.18 2.09 2.03 24.57 27.68 29.76 6 3.01 2.37 2.27 2.22 21.26 24.58 26.25 7 3.19 2.60 2.54 2.50 18.50 20.38 21.63 日最低
气温1 1.43 0.92 0.93 0.91 35.66 34.97 36.36 2 1.46 1.08 1.02 0.97 26.03 30.14 33.56 3 1.43 1.19 1.13 1.07 16.78 20.98 25.17 4 1.39 1.26 1.20 1.15 9.35 13.67 17.27 5 1.44 1.37 1.31 1.24 4.86 9.03 13.89 6 1.55 1.49 1.41 1.35 3.87 9.03 12.90 7 1.66 1.67 1.55 1.49 -0.60 6.63 10.24
下载: 导出CSV
-
[1] 张庆奎, 寿绍文, 陆汉城.卡尔曼滤波方法在极端温度预报中的应用.科技信息, 2008(35):26-27. doi: 10.3969/j.issn.1673-1328.2008.35.026 [2] 吴君, 裴洪芹, 石莹, 等.基于数值预报产品的地面气温BP-MOS预报方法.气象科学, 2007, 27(4):430-435. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKX200704013.htm [3] 赵声蓉.多模式温度集成预报.应用气象学报, 2006, 17(1):52-58. doi: 10.11898/1001-7313.20060109 [4] 车钦, 赵声蓉, 范广洲.华北地区极端温度MOS预报的季节划分.应用气象学报, 2011, 22(4):429-436. doi: 10.11898/1001-7313.20110405 [5] 周慧, 崔应杰, 胡江凯, 等.T639模式对2008年长江流域重大灾害性降水天气过程预报性能的检验分析.气象, 2010, 36(9):60-67. doi: 10.7519/j.issn.1000-0526.2010.09.010 [6] 熊秋芬.GRAPES_Meso模式的降水格点检验和站点检验分析.气象, 2011, 37(2):185-193. doi: 10.7519/j.issn.1000-0526.2011.02.008 [7] 周兵, 赵翠光, 赵声蓉.多模式集合预报技术及其分析与检验.应用气象学报, 2006, 17(增刊Ⅰ):104-109. http://www.cnki.com.cn/Article/CJFDTOTAL-YYQX2006S1014.htm [8] 赵声蓉, 裴海瑛.客观定量预报中降水的预处理.应用气象学报, 2007, 18(1):21-28. doi: 10.11898/1001-7313.20070104 [9] 赵声蓉, 赵翠光, 赵瑞霞, 等.我国精细化客观气象要素预报进展.气象科技进展, 2012, 2(5):12-21. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKZ201205005.htm [10] 刘还珠, 赵声蓉, 陆志善, 等.国家气象中心气象要素的客观预报——MOS系统.应用气象学报, 2004, 15(2):181-191. http://qikan.camscma.cn/jams/ch/reader/view_abstract.aspx?file_no=20040223&flag=1 [11] 马清, 龚建东, 李莉, 等.超级集合预报的误差订正与集成研究.气象, 2008, 34(3):42-48. doi: 10.7519/j.issn.1000-0526.2008.03.007 [12] 李佰平, 智协飞.ECMWF模式地面气温预报的四种误差订正方法的比较研究.气象, 2012, 38(8):897-902. doi: 10.7519/j.issn.1000-0526.2012.08.001 [13] 智协飞, 伍清, 白永清, 等.IPCC-AR4模式资料的地面气温超级集合预测.气象科学, 2010, 30(5):708-714. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKX201005018.htm [14] 林春泽, 智协飞, 韩艳, 等.基于TIGGE资料的地面气温多模式超级集合预报.应用气象学报, 2009, 20(6):706-712. doi: 10.11898/1001-7313.20090608 [15] 范丽军, 符淙斌, 陈德亮.统计降尺度法对华北地区未来区域气温变化情景的预估.大气科学, 2007, 31(5):887-897. http://www.cnki.com.cn/Article/CJFDTOTAL-DQXK200705011.htm [16] 王敏, 李晓莉, 范广洲, 等.区域集合预报系统2 m温度预报的校准技术.应用气象学报, 2012, 23(4):395-401. doi: 10.11898/1001-7313.20120402 [17] 李刚, 吴春燕, 肖若.地面气温的概率预报试验.气象科技, 2015, 43(1):97-102. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKJ201501018.htm [18] 陈晓龙, 智协飞.基于TIGGE资料的北半球地面气温预报的统计降尺度研究.大气科学学报, 2014, 37(3):268-275. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX201403002.htm [19] 邱学兴, 王东勇, 朱红芳.乡镇精细化最高最低气温预报方法研究.气象与环境学报, 2013, 29(3):92-96. http://www.cnki.com.cn/Article/CJFDTOTAL-LNQX201303016.htm [20] 智协飞, 林春泽, 白永清, 等.北半球中纬度地区地面气温的超级集合预报.气象科学, 2009, 29(5):569-574. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKX200905002.htm [21] 智协飞, 李刚, 彭婷.基于贝叶斯理论的单站地面气温的概率预报研究.大气科学学报, 2014, 37(6):740-748. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX201406008.htm [22] 崔慧慧, 智协飞.基于TIGGE资料的地面气温延伸期多模式集成预报.大气科学学报, 2013, 36(2):165-173. http://www.cnki.com.cn/Article/CJFDTOTAL-NJQX201302006.htm [23] 刘建国, 谢正辉, 赵琳娜, 等.基于TIGGE多模式集合的24小时气温BMA概率预报.大气科学, 2013, 37(1):43-53. doi: 10.3878/j.issn.1006-9895.2012.11232 [24] 罗菊英, 周建山, 闫永财.基于数值预报及上级指导产品的本地气温MOS预报方法.气象科技, 2014, 42(3):443-450. http://www.cnki.com.cn/Article/CJFDTOTAL-QXKJ201403016.htm [25] 吴振玲, 潘璇, 董昊, 等.天津市多模式气温集成预报方法.应用气象学报, 2014, 25(3):293-301. doi: 10.11898/1001-7313.20140305 -
计量
- 摘要浏览量: 6205
- HTML全文浏览量: 1681
- PDF下载量: 883
- 被引次数: 0
