CDF-T方法在站点尺度日降水预估中的应用

基于1961-2015年上海降水观测数据和8个全球气候模式GCMs模拟的日降水量数据,采用累计概率分布函数构建转换模型CDF-T建立了站点尺度日降水量的统计降尺度模型。结果表明,降尺度模型显著改善了GCMs对降水日数偏多、降水强度偏低和降水量偏少的模拟结果。与利用全年日降水序列建模结果相比,利用汛期日降水序列建模更好地刻画了汛期降水的累计概率分布曲线,同时提高了汛期总降水量、降水强度和年平均暴雨日数、暴雨量、暴雨强度的均值和变化趋势的降尺度效果。模型对较长年份的暴雨重现期订正效果更佳。与当代(2006-2015年)气候相比,2016-2095年上海降水呈现以下特征:全年和汛期总降水量和降水强度增加,降水日数减少,未来可能出现更多的旱涝年;汛期降水极端性增强,暴雨降水均值和极端值均增加;50年以上重现期的年最大日降水量未来呈前40年减少后40年增加的变化。CDF-T模型为站点尺度气候变化影响评估和未来预估提供降尺度技术和基础气候数据。     

自组织映射神经网络(SOM)降尺度方法对江淮流域逐日降水量的模拟评估

利用1961~2002年ERA-40逐日再分析资料和江淮流域56个台站逐日观测降水量资料,引入基于自组织映射神经网络(Self-Organizing Maps,简称SOM)的统计降尺度方法,对江淮流域夏季(6~8月)逐日降水量进行统计建模与验证,以考察SOM对中国东部季风降水和极端降水的统计降尺度模拟能力。结果表明,SOM通过建立主要天气型与局地降水的条件转换关系,能够再现与观测一致的日降水量概率分布特征,所有台站基于概率分布函数的Brier评分(Brier Score)均近似为0,显著性评分(Significance Score)全部在0.8以上;模拟的多年平均降水日数、中雨日数、夏季总降水量、日降水强度、极端降水阈值和极端降水贡献率区域平均的偏差都低于11%;并且能够在一定程度上模拟出江淮流域夏季降水的时间变率。进一步将SOM降尺度模型应用到BCCCSM1.1(m)模式当前气候情景下,评估其对耦合模式模拟结果的改善能力。发现降尺度显著改善了模式对极端降水模拟偏弱的缺陷,对不同降水指数的模拟较BCC-CSM1.1(m)模式显著提高,降尺度后所有台站6个降水指数的相对误差百分率基本在20%以内,偏差比降尺度前减小了40%~60%;降尺度后6个降水指数气候场的空间相关系数提高到0.9,相对标准差均接近1.0,并且均方根误差在0.5以下。表明SOM降尺度方法显著提高日降水概率分布,特别是概率分布曲线尾部特征的模拟能力,极大改善了模式对极端降水场的模拟能力,为提高未来预估能力提供了基础。     

非齐次隐马尔可夫降尺度方法对江淮流域夏季逐日降水的模拟及其评估

引入非齐次隐马尔可夫模型(Nonhomogeneous hidden Markov model,NHMM)统计降尺度方法,利用1961—2002年江淮流域夏季逐日降水资料、欧洲中期天气预报中心(ECMWF)的ERA-40再分析资料建立模型,检验其对东部季风区(以江淮流域为代表)夏季日降水的模拟能力,并对比BCC-CSM1.1(m)模式NHMM降尺度前后的模拟效果。结果表明,NHMM降尺度方法通过建立降水概率分布态间转移参数与大尺度环流变量的联系,对江淮流域逐日降水量具有较好的降尺度效果。模拟的各站日降水量概率分布函数(PDF)曲线与观测非常接近,布赖尔评分(Brier Score,S_B)均小于0.11%,显著性评分(Significance Score,Ss)均大于0.84;夏季总降水量、降水日数、中雨日数、降水强度和95%分位降水量指数的多年平均场偏差百分率绝对值低于10%,前3个指数的空间相关系数高于0.9;该方法对各降水指数的年际变率也有一定的模拟能力,模拟得到的各指数的区域平均年际序列与观测序列的相关系数为0.62—0.87。对BCC-CSM1.1(m)模式的模拟结果进行降尺度后,SB较降尺度前平均减小0.57%,Ss平均增大0.23,皆表明降尺度后的概率分布函数曲线更接近于观测;各降水指数在多数台站的偏差百分率绝对值由大于40%降至10%以内,空间相关系数普遍提高至0.8以上。NHMM降尺度方法能够有效提高BCC-CSM1.1(m)模式对江淮流域夏季日降水的模拟能力,相对气候模式具有显著的"增值",未来可进一步利用该方法进行气候变暖背景下的日降水变化预估。     

气候变化影响下的流域水循环:以长江和黄河典型流域为例

在对IPCC提供的多种大气环流模式(GCMs)适用性评估的基础上,采用SDSM和ASD统计降尺度模型生成未来气候变化情景,驱动分布式水文模型VIC和SWAT,分别对长江和黄河典型流域进行分布式水文模拟,定量探讨气候变化对流域水循环的影响。结果表明,适应性评估可以有效降低GCMs选择的不确定性,统计降尺度方法能够明显改善全球气候模式降水和气温输出结果。与基准期(1961—1990年)相比,未来时期(2046—2065年和2081—2100年)长江下游太湖流域径流量呈现微弱减少趋势,但汛期东南部径流量增加显著,而黄河上游流域径流量则呈下降趋势。研究结果可为开展我国各大流域适应气候变化研究提供一定的参考依据。     

全球气候模式对宁夏区域未来气候变化的情景模拟分析

利用多个全球气候模式(GCM)的情景模拟结果分析只考虑温室气体效应的IS92a GG情景和同时考虑温室气体效应和硫化物气溶胶辐射效应的IS92aGS情景以及SRESA2、B2情景下宁夏区域21世纪地面气温和降水量的可能变化,并进行不确定性分析。气候基准时段(1961~1990年)模拟结果与观测资料的对比分析表明,GCM对宁夏气候具有一定的模拟能力;整体上讲,GCM对地面气温的模拟值偏低,对降水量的模拟值偏高,其中ECHAM4和HadCM3对宁夏基准时段地面气温和降水量的模拟结果与观测比较接近。各GCM模拟值的平均结果显示,4种温室气体排放情景下21世纪宁夏区域气温持续升高,至21世纪末宁夏升温幅度可达4~6℃,与全国平均的增温幅度大致相当;与升温趋势相应的是降水量的增加,但降水变化呈现出很大的波动性,至21世纪末宁夏的降水变化幅度可达10%~40%。各个GCM模拟的宁夏气候变化的总趋势是一致的,但各模式在不同情景下模拟结果的差异很大,存在较大的不确定性。     

基于ASD统计降尺度的雅鲁藏布江流域未来气候变化情景

青藏高原对东亚、南亚甚至全球的气候和水文循环有重要影响,模拟和分析青藏高原未来气候状况对研究东南亚区域生态、气候、水资源格局与演变趋势具有重要意义。利用ASD统计降尺度方法,对MIROC3.2_medres模式输出的降水和气温进行降尺度,并与ERA-40再分析资料进行了对比分析和评价,在此基础上分析了2046-2065年与2081-2100年三种情景下的气候变化情况。结果表明,ASD统计降尺度模型可以较好地将雅鲁藏布江流域GCM大尺度降水和气温数据降尺度到站点尺度,气温的解释方差都在90%以上,降水的解释方差也达到12%~27%;雅鲁藏布江流域未来降水年际变化不明显,年变化趋势多数小于5%;降水年内分配将更加集中,秋、冬、春季的降水减少趋势明显,最大降幅达55.58%,夏季降水显著增加,最大增幅达到30.44%;雅鲁藏布江流域未来将显著增温,21世纪中叶增温幅度为1.60~2.12℃,21世纪末期增温幅度达2.34~3.69℃;在降水与气温的双重影响下,流域水资源问题可能会变得更加严峻。     

基于分位数映射法的黑河上游气候模式降水误差订正

区域气候模式降水弥补了高寒山区气象站点稀少的缺陷,是水文模拟的重要驱动变量。然而,高寒山区模式输出降水的总量和频率都存在较大不确定性。因此,改进了用于降水频率纠正的分位数映射法(Quantile Mapping,QM),对中尺度数值预报模式(Weather Research and Forecasting model,WRF)模拟的黑河上游日降水输出数据进行误差订正。选取第95分位和第98分位降水量为阈值,选择2004-2009年为建模时段,2010-2013年为验证时段,使用分段拟合的方法建立传递函数,侧重于对极端降水进行单独订正。研究结果表明:该方法不仅对降水空间分布有明显的改善,对极端降水也有很好的订正效果。订正前模式模拟日降水与台站之间的均方根误差为3.41 mm·d^-1,绝对偏差为115.67 mm·y^-1,订正后均方根误差减少为3.11 mm·d^-1,绝对偏差有明显改善,为60.3 mm·y^-1。订正后流域内年降水空间分布更加合理,年降水量也更接近于观测降水插值结果,其空间相关系数由0.74改善为0.94。春、夏季订正效果优于秋、冬季,其中夏季订正效果较为明显,订正前降水偏差百分比在-0.1~0.1以内的区域面积仅占流域总面积的28%,而订正后占比增加至66%。同时,该方法对极端降水有较好的订正效果,减小了日降水强度(SDII)和极强降水量(R99p)的模拟偏差,订正后的第95分位模拟降水与观测降水插值的相关系数由0.15提高到0.48。本研究为站点稀少的黑河上游提供了一种更有效的误差订正方案,有利于为寒区水文研究获取更精确的降水数据。     

基于多模式降水量预报的浙江省统计降尺度研究

基于欧洲中期天气预报中心(ECMWF)、美国全球集合预报系统(GFS)、日本气象厅(JMA)3个中心3~96 h预报时效的降水量预报资料,以及浙江省1957个高密度的自动站观测资料,对数值模式预报结果进行统计降尺度处理。首先利用线性回归方法对插值后的预报结果进行降尺度订正,然后把3个数值预报的降尺度结果进行消除偏差集合平均,最后得到多模式集成的降水量预报场。结果表明,统计降尺度订正后的预报结果比直接插值更加准确,多模式集成的预报效果优于单模式结果,其改进效果随预报时效的延长减小。     

基于概率匹配的西南区域模式定量降水订正试验

利用2014—2015年5—10月地面观测降水资料和同时段的西南区域模式降水预报资料,基于概率匹配方法,采取分区及点对点匹配两种方案对2016年6—8月降水集中时段逐12h累积降水进行订正试验。结果表明:(1)订正后的模式预报相比订正前而言,平均(绝对)误差有所减小,降水落区的范围和平均强度与实况更加接近;(2)量级偏差越大,运用该方法的订正效果越好,夜间降水订正效果优于白天;(3)分区统计方案对模式系统性偏差的订正效果优于点对点方案,合理的区域划分增加统计样本量可以提高订正效果。     

遥感估算降水在西藏高原中的应用研究

采用遥感估算降水模型RFE 2.0(Rainfall Estimation Algorithm Version 2)模拟了2009年西藏高原的区域降水,并结合该地区气象站降水观测资料分别从日、月、年尺度上评价了该模型在西藏高原降水估算中的适用性,最后通过系数校正分析了2009年8月西藏高原降水量和年降水量的分布格局。结果表明,RFE2.0模型日降水量模拟值与观测值的相关系数在0.40以上的测站占46%,变化趋势较一致,但在日降水量较小时(接近零)模拟结果不稳定,在降水量较大时(>15mm)模拟结果一般会偏低;月平均降水量模拟结果与观测结果的相关系数在0.80以上的测站占62%,模拟结果较好地反映了观测结果的变化趋势,但个别月份的模拟结果会出现偏差。雨季降水量的模拟结果明显好于干季,为进一步提高模拟精度,确定雨季校正系数为1.133,干季校正系数为1.265;年尺度上降水量的模拟值与观测值的相关系数为0.368(P=0.026)。整体来看,遥感估算降水模型(RFE2.0)模拟的西藏高原降水结果较好,可为西藏高原降水模拟提供借鉴和参考。     

江淮梅雨期降水南北反位相分布与大气准双周振荡

利用1954—2005年中国740站逐日降水资料和NCEP/NCAR再分析资料,分析江淮梅雨期降水的南北反位相分布(Anti-Phase Distribution:APD)和大气准双周振荡(Qusi-Biweekly Oscillation:QBWO)之间的关系,诊断结果表明:(1)1954—2005年江淮梅雨APD共有16年较为显著。从1990年代开始,梅雨APD显著增强,并且表现出显著的2年和4～6年振荡周期。在降水较多的区域,降水的准双周振荡往往也较强。(2)梅雨期降水APD和中国东部地区降水的南北变动同属一个位相,中国东部地区的南涝北旱或者南旱北涝在很大程度上可以由江淮地区降水的分布类型来说明。(3)"南旱北涝"年,准双周滤波的整层水汽通量能够传播到30°N以北,同时存在强烈的水汽通量辐合从中高纬度向南传播到江淮流域。而在"南涝北旱"年,准双周的水汽输送所能到达的纬度明显偏南,来自中高纬度向南传播的水汽通量辐合也不显著。(4)"南旱北涝"年降水正位相,西北太平洋副热带高压(简称西太副高)脊线位于22°N以北,850 hPa低频反气旋的位置相对于"南涝北旱"年偏北,调节西太副高进入南海位置也偏北。南海和江淮上空的准双周垂直速度异常位置整体偏北,第3位相的上升运动和第7位相的下沉运动都位于30°N以北,南海的垂直速度异常也主要位于南海北部,而在"南涝北旱"年,准双周垂直速度异常的分布偏南。     

Modelling daily temperature extremes: recent climate and future changes over Europe

Probability distributions of daily maximum and minimum temperatures in a suite of ten RCMs are investigated for (1) biases compared to observations in the present day climate and (2) climate change signals compared to the simulated present day climate. The simulated inter-model differences and climate changes are also compared to the observed natural variability as reflected in some very long instrumental records. All models have been forced with driving conditions from the same global model and run for both a control period and a future scenario period following the A2 emission scenario from IPCC. We find that the bias in the fifth percentile of daily minimum temperatures in winter and at the 95th percentile of daily maximum temperature during summer is smaller than 3 (±5°C) when averaged over most (all) European sub-regions. The simulated changes in extreme temperatures both in summer and winter are larger than changes in the median for large areas. Differences between models are larger for the extremes than for mean temperatures. A comparison with historical data shows that the spread in model predicted changes in extreme temperatures is larger than the natural variability during the last centuries.     

Statistical modeling of CMIP5 projected changes in extreme wet spells over China in the late 21st century

The observed intensity, frequency, and duration (IFD) of summer wet spells, defined here as extreme events with one or more consecutive days in which daily precipitation exceeds a given threshold (the 95th percentile), and their future changes in RCP4.5 and RCP8.5 in the late 21st century over China, are investigated by using the wet spell model (WSM) and by extending the point process approach to extreme value analysis. Wet spell intensity is modeled by a conditional generalized Pareto distribution, frequency by a Poisson distribution, and duration by a geometric distribution, respectively. The WSM is able to realistically model summer extreme rainfall spells during 1961–2005, as verified with observations at 553 stations throughout China. To minimize the impact of systematic biases over China in the global climate models (GCMs) of the Coupled Model Intercomparison Project Phase 5 (CMIP5), five best GCMs are selected based on their performance to reproduce observed wet spell IFD and average precipitation during the historical period. Furthermore, a quantile–quantile scaling correction procedure is proposed and applied to produce ensemble projections of wet spell IFD and corresponding probability distributions. The results show that in the late 21st century, most of China will experience more extreme rainfall and less low-intensity rainfall. The intensity and frequency of wet spells are projected to increase considerably, while the duration of wet spells will increase but to a much less extent. The IFD changes in RCP8.5 are in general much larger than those in RCP4.5.     

Assessing uncertainties in climate change impact analyses on the river flow regimes in the UK. Part 1: baseline climate

Assessing future climate and its potential implications on river flows is a key challenge facing water resource planners. Sound, scientifically-based advice to decision makers also needs to incorporate information on the uncertainty in the results. Moreover, existing bias in the reproduction of the ‘current’ (or baseline) river flow regime is likely to transfer to the simulations of flow in future time horizons, and it is thus critical to undertake baseline flow assessment while undertaking future impacts studies. This paper investigates the three main sources of uncertainty surrounding climate change impact studies on river flows: uncertainty in GCMs, in downscaling techniques and in hydrological modelling. The study looked at four British catchments’ flow series simulated by a lumped conceptual rainfall–runoff model with observed and GCM-derived rainfall series representative of the baseline time horizon (1961–1990). A block-resample technique was used to assess climate variability, either from observed records (natural variability) or reproduced by GCMs. Variations in mean monthly flows due to hydrological model uncertainty from different model structures or model parameters were also evaluated. Three GCMs (HadCM3, CCGCM2, and CSIRO-mk2) and two downscaling techniques (SDSM and HadRM3) were considered. Results showed that for all four catchments, GCM uncertainty is generally larger than downscaling uncertainty, and both are consistently greater than uncertainty from hydrological modelling or natural variability. No GCM or downscaling technique was found to be significantly better or to have a systematic bias smaller than the others. This highlights the need to consider more than one GCM and downscaling technique in impact studies, and to assess the bias they introduce when modelling river flows.     

The Application of a Meteo-hydrological Forecasting System with Rainfall Bias Correction in a Small and Medium-sized Catchment

Meteo-hydrological forecasting models are an effective way to generate high-resolution gridded rainfall data for water source research and flood forecast. The quality of rainfall data in terms of both intensity and distribution is very important for establishing a reliable meteo-hydrological forecasting model. To improve the accuracy of rainfall data, the successive correction method is introduced to correct the bias of rainfall, and a meteo-hydrological forecasting model based on WRF and WRF-Hydro is applied for streamflow forecast over the Zhanghe River catchment in China. The performance of WRF rainfall is compared with the China Meteorological Administration Multi-source Precipitation Analysis System (CMPAS), and the simulated streamflow from the model is further studied. It shows that the corrected WRF rainfall is more similar to the CMPAS in both temporal and spatial distribution than the original WRF rainfall. By contrast, the statistical metrics of the corrected WRF rainfall are better. When the corrected WRF rainfall is used to drive the WRF-Hydro model, the simulated streamflow of most events is significantly improved in both hydrographs and volume than that of using the original WRF rainfall. Among the studied events, the largest improvement of the NSE is from -0.68 to 0.67. It proves that correcting the bias of WRF rainfall with the successive correction method can greatly improve the performance of streamflow forecast. In general, the WRF / WRF-Hydro meteo-hydrological forecasting model based on the successive correction method has the potential to provide better streamflow forecast in the Zhanghe River catchment.     

基于NCEP-GEFS回算资料的我国极端温度变化特征研究

利用美国NCEP全球集合预报系统(GEFS)历史回算资料和中国均一化格点观测数据,分析了我国近30 a来极端温度变化特征,重点考察了该模式预报系统对这一变化特征的刻画性能。通过估算格点观测和模式资料中2 m温度的历史气候百分位,分析了我国冬夏两季极端温度的气候特征以及极端温度日数的气候分布和多年变化趋势。结果表明,我国冬季极端低温和夏季极端高温的空间分布表现出较强区域性特征:东北、华北和青藏高原区域冬季极端低温的百分位阈值对应的温度较低,而华南、西北和长江流域夏季极端高温的阈值温度则较高;近30 a来我国夏季平均温度和极端高温日数几乎都呈现上升趋势,冬季平均温度则在我国大部分区域呈上升趋势、西北和东北部分地区呈下降趋势,相应地冬季极端低温日数在大部分区域呈下降趋势、仅在西北、东北和华南部分地区略有上升。NCEP-GEFS回算资料能较好地再现我国冬夏两季平均气温、冬季极端低温和夏季极端高温日数的气候趋势和年际变化,但在各区域都有不同程度的冷偏差,冬季偏差明显大于夏季,并随着预报时长的增加,冬季冷偏差逐渐增强,而夏季冷偏差则逐渐减弱。因此,本文建议采用基于百分位阈值的相对极端性定义,可自动修正模式分析场和预报场中的系统性偏差。     

中国主要河流流域极端降水变化特征

利用中国1956-2008年逐日降水量资料,以全国主要河流流域为研究区域,分析了年最大日降水量、年暴雨（日降水量≥50.0 mm）日数的多年平均状况及长期变化趋势。分析表明,近53年,全国平均年最大日降水量没有明显的线性变化趋势,但全国范围内多数气象站点年最大日降水量呈现出增加趋势,并存在南方流域增加、北方流域减少的变化趋势,这种变化特征在2001年以来表现更加突出。全国平均年暴雨日数呈不显著的增多趋势,20世纪90年代最多,70年代最少。空间上,我国南北方流域年暴雨日数呈现相反的变化特征,南方流域多呈上升趋势,北方流域呈减少趋势。     

Improved atmospheric circulation over Europe by the new generation of CMIP6 earth system models

Global Climate Models (GCMs) generally exhibit significant biases in the representation of large-scale atmospheric circulation. Even after a sensible bias adjustment these errors remain and are inherited to some extent by the derived downscaling products, impairing the credibility of future regional projections. In this study we perform a process-based evaluation of state-of-the-art GCMs from CMIP5 and CMIP6, with a focus on the simulation of the synoptic climatological patterns having a most prominent effect on the European climate. To this aim, we use the Lamb Weather Type Classification (LWT, Lamb British isles weather types and a register of the daily sequence 736 of circulation patterns 1861-1971. METEOROL OFF, GEOPHYS MEM; 737 GB; DA 1972; NO 116; PP 1-85; BIBL 2P1/2, 1972), a subjective classification of circulation weather types constructed upon historical simulations of daily mean sea level pressure. Observational uncertainty has been taken into account by considering four different reanalysis products of varying characteristics. Our evaluation unveils an overall improvement of salient atmospheric circulation features consistent across observational references, although this is uneven across models and large frequency biases still remain for the main LWTs. Some CMIP6 models attain similar or even worse results than their CMIP5 counterparts, although in most cases consistent improvements have been found, demonstrating the ability of the new models to better capture key synoptic conditions. In light of the large differences found across models, we advocate for a careful selection of driving GCMs in downscaling experiments with a special focus on large-scale atmospheric circulation aspects.

     

Searching for Added Value in Simulating Climate Extremes with a High-Resolution Regional Climate Model over Western Canada. II: Basin-Scale Results

We evaluate the capacity of a regional climate model to simulate the statistics of extreme events, and also examine the effect of differing horizontal resolution, at the scale of individual hydrological basins in the topographically complex province of British Columbia, Canada. Two climate simulations of western Canada (WCan) were conducted with the Canadian Regional Climate Model (version 4) at 15 (CRCM15) and 45?km (CRCM45) horizontal resolution driven at the lateral boundaries by global reanalysis over the period 1973–1995. The simulations were evaluated with ANUSPLIN, a daily observational gridded surface temperature and precipitation product and with meteorological data recorded at 28 stations within the upper Peace, Nechako, and upper Columbia River basins. In this work, we focus largely on a comparison of the skill of each model configuration in simulating the 90th percentile of daily precipitation (PR90). The companion paper describes the results for a wider range of temperature and precipitation extremes over the entire WCan domain.

Over all three watersheds, both simulations exhibit cold biases compared with observations, with the bias exacerbated at higher resolution. Although both simulations generally display wet biases in median precipitation, CRCM15 features a reduced bias in PR90 in all three basins in summer and throughout the year in the upper Columbia River basin. However, the higher resolution model is inferior to CRCM45 with respect to rarer heavy precipitation events and also displays high spatial variability and lower spatial correlations with ANUSPLIN compared with the coarser resolution model. A reduction in the range of PR90 biases over the upper Columbia basin is noted when the 15?km results are averaged to the 45?km grid. This improvement is partly attributable to the averaging of errors between different elevation data used in the gridded observations and CRCM, but the sensitivity of CRCM15 to resolved topography is also clear from spatial maps of seasonal extremes. At the station scale, modest but systematic reductions in the bias of PR90 relative to ANUSPLIN are again found when the CRCM15 results are averaged to the 45?km grid. Furthermore, the annual cycle of inter-station spatial variance in the upper Columbia River basin is well reproduced by CRCM15 but not by ANUSPLIN or CRCM45. The former result highlights the beneficial effect of spatial averaging of small-scale climate variability, whereas the latter is evidently a demonstration of the added value at high resolution vis-à-vis the improved simulation of precipitation at the resolution limit of the model.