【作者】 王冀；

【导师】 江志红；

【作者基本信息】 南京信息工程大学 ， 气候系统与全球变化， 2008， 博士

【摘要】 本文利用1961-2006年中国中东部台站极端气温指数,分析了近46年中国中东部极端气温事件时空情况;应用IPCC-AR4提供GCM模拟极端气温指数结果对我国模拟能力评估并对未来21世纪不同排放情景下极端事件变化情况进行预估,同时利用统计(SDSM)和动力(LMDZ)方法对中国中东部极端气温变化进行降尺度模拟和评估,并采用评估效果较好的LMDZ区域降尺度方法预估中国中东部极端气温指数在21世纪中期变化情况。得到如下结论:1、IPCC-AR4提供的7个海气耦合模式对极端气温指数都具有一定的模拟能力,模式平均的模拟效果比单独用某一个模式要好。在年际的变化特征方面,模式平均的暖夜指数和霜冻日数相关系数最大分别为,0.74、0.68;生长季指数和温度年较差与观测数据的相关性较差。空间分布的评估结果表明,在全国范围内,以110°E为界,东部地区模拟效果较好,西部地区的模拟效果较差,对青藏高原地区的模拟效果最差。在极端气温指数中,模拟效果最好的是温度年较差,暖夜指数最差。综合评价表明GFDL-CM2.0和MIROC3.2(hires)对大多数指数的模拟效果都比较好。在未来21世纪A2情景下,霜冻日数、温度年较差呈减少趋势,其他气温指数均呈明显的增加趋势,热浪指数的增加趋势最明显,趋势系数达到90.2d/100a。从空间尺度变化发现,极端气温指数基本上是由北向南变化率逐渐增大,其中西北的极端指数变化率高于东北。2、中国中东部地区近46年以来夏日天数、年最低气温、年最低气温、暖夜指数、暖日指数和热浪指数均呈上升趋势,而霜冻日数、冷夜指数和冷日指数均呈下降趋势。区域平均的年最高气温、热浪指数变化趋势最小,分别为-0.063℃/10a、0.08%/10a,区域平均的霜冻日数变化趋势最明显,为-3.3d/10a。在年代际变化方面,表现出极端最低气温事件明显减少,极端高温事件有增多的趋势,但幅度要小于极端低温事件。极端气温事件在冬季和夏季变化的趋势明显,在春季和秋季变化幅度较小。在极端气温空间分布方面,霜冻日数的下降趋势最为明显,其中江苏和安徽地区变化最大,中心值达到-8天/10a。年最高气温、冷日指数、暖日指数的变化较小,在全区范围内通过显著性检验的主要为陕西和江浙沿海地区。3、SDSM-had同所嵌套的Hadcm3相比,在数值上更加接近实况,SDSM-had模拟区域平均极端气温误差均小于Hadcm3的模拟结果。在年际变化(时间序列相关系数)和空间分布特征(空间相关系数)方面SDSM-had模拟精度均强于Hadcm3,其中空间相关系数最好,均超过0.98。在SDSM-had模拟能力方面,SDSM-had模拟均方差小于观测值,这表明了SDSM-had极端值出现的频率小于实测资料。SDSM-had对武汉和南京的模拟效果较好。SDSM-had对中国中东部地区极端最高和最低气温空间分布具备很好的模拟能力,但存在明显的系统误差,其中SDSM-had对冬季最低气温的模拟(误差均值2℃)好于夏季最高气温(误差均值-2.8℃)。30年重现期极端气温结果表明,SDSM-had对重现期最低气温的模拟效果很差,误差平均在6℃以上,误差最大值为12℃。4、LMDZ区域气候模式的模拟效果要明显好于全球模式的模拟结果。区域气候模式模拟中国中东部地区季节和逐月平均最高和最低气温的误差均小于全球模式。变网格区域气候模式对极端气温的年际变化和空间分布均高于LMDZ-gcm,最高气温时间相关系数超过0.60,最低气温空间相关系数在0.90以上。变网格气候模式模拟代表站的结果表明,LMDZ-era40对极端气温均值、均方差和时间相关系数模拟效果好于LMDZ-reg,对武汉和南京模拟的效果最好。空间分布特征评估发现,两种方案对江苏、安徽、湖北有较好的模拟效果,而陕西是模拟效果最差的地区,其次是闽浙地区,LMDZ-reg比LMi)Z-era40与观测值误差更大。两种方案对30年重现期极端气温模拟有基本相同的空间分布特征,LMDZ-reg的模拟30年重现期最高气温误差相对较大。5、在未来21世纪中期A2情景下,中国中东地区平均的各月极端最高和最低气温呈增加趋势。最高气温增加最多的月份是2月,幅度达到2.81℃;增加最少的是5月,只有1.2℃。最低气温增加最多的月份是秋季11月份,增加最少的月份也发生在春季4月份,增加幅度为1.02℃。A2情景下未来21世纪中国中东部极端最低气温在各季节表现出由北向南变化一致减少的趋势,基本上(夏季除外)变化幅度最大的地区在江淮流域北部,变化幅度最小的地区一般在南部的广东和福建。而极端最高气温各季节则是在长江中下游流域存在变化的最大区域,而沿海的福建一带始终是变化最小的区域。30年重现期极端气温均有所升高,平均增加2℃左右,其中广东和福建地区存在重现期最高气温增加的高值区,中心值为7℃,在A2情景下,暖夜指数增加最显著,增幅最大的地区在广东、广西以及福建南部,增加幅度在85%以上,热浪指数变化最小。更多还原

【Abstract】 This paper has evaluated the simulation ability of seven coupled General Circulation Models (GCM) supplied by the Intergovernmental Panel on Climate Change’s 4th Assessment Report (IPCC-AR4) which are applied into climate simulations for the extreme temperature indices in China, followed by forecast on the changes in extreme events in the 21st century on different emissions scenarios. Then variations in the extreme temperature events over mid-east China during 1961-2006 are investigated by the extreme temperature indices therein. Last, SDSM and LMDZ down scaling methods are used in the simulation and evaluation of extreme temperature over mid-east China. As the LMDZ method has better results, it is utilized to forecast the variations of the extreme temperature indices over mid-east China in the 21st century.The main results are as follows:1. Based on the 1961～2000 extreme temperature observations in China, we have evaluated the simulations and predictions of products such as frost days (FD), growing season length (GSL), extreme temperature range (ETR), warm nights (TN90), and heat wave duration index (HWDI) from seven coupled General Circulation Models (GCM) supplied by the Intergovernmental Panel on Climate Change’s 4th Assessment Report (IPCC-AR4). Results indicate that all the seven models are able to simulate the extreme temperature indices of China in a certain degree, while their ensemble acts better than any single one. For each index, most of the models can present linear trends of the same positive/negative signs as the observations but for weaker intensities. The simulation effects are different on a nationwide basis, with 110°N as the division, east (west) of which the effects are better (worse) and the worst over the Qinghai -Tibetan plateau. The predictions for the 21st century on emissions scenarios show that except decreases in the FD and ETR, other indices display significant increasing trend, especially for the indices of notable extreme climate such as HWDI and TN90, indicating that the temperature-related climate is moving towards the extreme. Under different kinds of emissions scenarios as SRES A2, SRES A1B, and SRES B1, the extreme temperature indices are always increasing (decreasing) consistently in the 21st century, with decrease (increase) in the FD and ETR (GSL, HWDI and TN90) . In the high emissions scenarios (A2), trends of the indexes are most significant, with the least in the low emissions scenarios (B1). Among the five indices, the largest changing ranges are in the HWDI and TN90, FD and GSL as the second, and least in the ETR. Except the GSL, the distributions of other four indices’ trends are almost the same throughout China, with the variability rising gradually from north to south, and greater changes in northwest than northeast for indices of FD, ETR and HWDI, as different from the TN90 whose trends are more significant in southwest and south China.2. In the recent 46 years, the summer days (SU), yearly maximum and minimum temperature, warm nights (TN90), warm days (TX90) and heat wave duration index are increasing, as opposed to the FD, cold nights (TN10) and cold days (TX10), in which the yearly minimum temperature and TN90 have the largest rising trends. As respect to the inter-decadal variations, the extremely cold events are reduced significantly, while the extremely warm events are increased but with smaller ranges. The change of extreme temperature events are more distinct in the winter and summer rather than in the spring and autumn. As for the spatial distributions, the temperature indices have the positive or negative tendencies all over the study area, with national increasing (decreasing) trends in the SU, yearly minimum temperature , TN90 and HWDI (the FD , TN10 and TX10).3. SDSM-had simulations are better than the General Circulation Model Hadcm3 not only for the number of modeling products, but also for the simulated precision of the interannual variations (indicated by the connection coefficient between the temporal series) and the spatially distributions (indicated by the spatially connection coefficients). The SDSM-had can simulate the normal distribution of daily maximum and minimum temperatures for the stations, but with smaller frequency of the extreme values than the observations. The simulation effects of SDSM-had for the inter-decadal variations are better in the minimum temperature of winter than in the maximum temperatures of summer for the typical stations. The SDSM-had method are able to simulate the spatial distributions of extremely maximum/minimum temperatures over mid-east China and display the primary distribution characteristics of the real stations, but with significant systematic error, while the simulation effects are better for the minimum temperature in winter than maximum temperatures in summer. Extreme temperature return value evaluations indicate that the simulation capability of SDSM-had method are weak for the extreme minimum temperature.4. The simulations of LMDZ regional model are better than the GCM. The zoomed regional model can improve the simulating precision in the inter-decadal variations and spatial distributions of regional maximum and minimum temperatures. The zoomed climate model LMDZ-era40 and LMDZ-reg have the simulating capability for the maximum and minimum temperatures in mid-east China to some extent. Modeled results of typical stations show that the frequencies of the minimum temperature in winter and the maximum temperatures in summer are almost the same between the LMDZ-era40 and LMDZ-reg simulations, but better for the minimum temperature in winter than the maximum temperature in the summer. Mean square deviation, spatial/temporal connection coefficients indicate that the simulations of LMDZ-era40 are better than the LMDZ-reg. Evaluation of spatial distributions show that both the tests can simulate the main spatial distributions of the minimum temperature in winter and the maximum temperature in the summer with similar characteristics. Both are better over Jiangsu, Anhui, and Hubei provinces, while LMDZ-reg is less close to the observations than the LMDZ-era40. And both have the stable effects for extreme temperature return value evaluations, while the effects of minimum temperature return value evaluations are more stable than the maximum one.5. The LMDZ-reg simulations of the extreme temperature of mid-east China in the mid- 21st century under SRES A2 show that the monthly maximum and minimum temperatures may rise consistently but with different changing ranges. The maximum (minimum) temperatures increase greatest in the winter and autumn (winter and summer), least in the summer (spring). Generally, the frequency of extreme warm (cold) events is increasing (decreasing) notably. Variations in the GEV probability density functions are also show the significant warming of the extremely maximum and minimum temperatures, while the later increases more than the former. For annual and seasonal distributions (except the summer), the tendency of extremely minimum temperature are decreased from north to south in mid-east China, with the largest (smallest) change in the north Yangtze-Huaihe river valleys (Guangdong and Fujian). The annual and seasonal distributions of the extremely maximum temperature’s trends are most significant in the mid-lower Yangtze River valleys, and least in the coastal regions in Fujian. The TN90 indicates the largest changing range among the study indices; the FD is the second, and the HWDI as the least.更多还原