一种联合历史和实时资料的定量降水短时预报技术研究

本文提出一种基于百分位映射,使用实时和历史资料联合订正模式定量降水短时订正预报技术,并与仅使用历史资料或实时资料的订正效果进行对比。结果表明:三种方案均能有效订正模式原始降水预报偏差,提高0~12 h降水预报准确率。对于晴雨预报,采用联合订正方案,预报效果最优。在0~7 h预报时效内,仅采用实时资料的订正方案准确率明显优于仅采用历史资料的订正方案,在8~12 h预报时效内,后者准确率稍高。所有预报时效内,仅采用实时资料的订正方案降水预报范围较仅采用历史资料的方案略偏大。对于较强降水预报,采用联合订正方案准确率为三种方案最优,仅采用实时资料的方案预报准确率虽优于仅采用历史资料的方案,但预报范围及量级较实况明显偏大。     

Variational data assimilation in the tropics using precipitation data part I: Column model

Summary This paper describes initial effort in the development of a four-dimensional variational data assimilation (4D-Var) in the tropics using precipitation data derived from remote sensing. The method of 4D-Var using precipitation data is formulated, and modifications to the parameterization schemes of moist processes to remove zeroth-order discontinuities are described. Variational data assimilation experiments are carried out using a column model to investigate the problems caused by discontinuities in parameterization schemes and assess the impact of assimilating precipitation data in the tropics.It is found that variational data assimilation with discontinuous parameterization schemes exhibits large fluctuations during the minimization process, slow convergence rates, and large analysis errors. The fluctuations become much more serious when precipitation data is assimilated. Precipitation data is very useful to estimate divergence in the tropics, provided that the temporal resolution of the data is sufficiently high. However, its impact on the analysis of temperature and moisture is not clear in the column model assimilation experiments, possibly due to the absence of horizontal advection.     

基于频率匹配法的江西省ECMWF降水预报订正研究

利用2016—2019年ECMWF（European Centre for Medium-Range Weather Forecasts）降水预报资料和江西省93个国家气象站降水资料,基于频率匹配法进行降水偏差订正,采用4种方法统计降水频率对降水预报进行订正试验（不分区试验）,根据江西省汛期暴雨气候特征对汛期降水进行分区订正试验,并对典型强降水个例进行分析。结果表明: 频率匹配法降低了模式预报小雨的空报率和大雨、暴雨的漏报率,预报技巧改进明显。在4种降水频率统计方法中,准对称滑动平均法效果最好。分区试验对强降水的订正效果优于不分区试验,该试验对模式预报正技巧暴雨过程的订正能力大于无技巧过程。对于模式预报效果差（TS=0）、一般（0 < TS < 0.2）、好（TS≥0.2）的暴雨过程,分区试验改善的概率分别为40.8%、89.1%和65.3%。频率匹配分区订正后强降水面积更加接近实况,但强降水落区不能得到明显的改善。订正方法对模式预报强降水形态、位置与实况较接近的过程,效果较好。     

数值降水预报定量集成方法的业务应用

利用相似权重集成预报法对辽宁区域12个数值模式预报的降水量进行集成,并投入业务化应用。结果表明：通过对2009年5月1日至10月20日的24 h降水量预报进行TS评分,发现降水集成方法要优于12个集合成员的单个预报,同时也要优于简单的集合平均。试验表明,滞后时间和扩大圈数对集成预报效果影响很大,而单个的集合成员对集成预报效果影响较小。根据试验结果修改集成方法应用方案,按照不同降水量级和预报时效选择扩大的圈数,预报效果好于原方案,对大量级、长时效预报改进更明显,其中25 mm量级预报时效72 h的TS评分增加了20%或以上,具有实际应用价值。     

漂移克里金方法在雷达和雨量计联合估测降水中的应用

文中介绍了一种新的融合雷达和雨量计数据开展定量估测降水研究的空间信息统计学方法-Kriging with externaldrift(KED)方法.该方法能很好地融合高精度、低时空分辨率的雨量计数据和低精度、高时空分辨率的雷达数据进行插值.通过变异函数描述降水场的空间结构信息,能够充分利用数据间的空间相关性,来改进估测精度和提高处理速度.利用其优良的数学特性,以期在定量估测降水业务研究上进行新的探索和尝试.选用湖南省有代表意义的3次降水过程资料,通过雷达直接估测降水(RAD)、变分校准(VAR)以及KED3种方法,分别与雨量计测量值进行对比分析,选用代表站进行交叉验证结果均表明:RAD的均方差、绝对误差、相对误差最大,VAR次之,而KED最小.KED估测的结果与雨量计测量降水最为接近,估测效果最好;3种方法与雨量计实测值计算一定范围的误差频率,KED估测值具有最小的均方差和最小的标准差,且误差分布相对集中在0值附近,斜度和峰度最佳,试验证明该方法不仅能提高降水估测精度,且优于其他方法,VAR均方差次之,RAD均方差效果相对较差.联合雷达、雨量计估测降水的实质是把雷达估测值与雨量计测量的结果相融合,以雨量计来校准雷达估测值,保留了雷达探测剑降水的中、小尺度精细特征.校准后的雨量场数值接近雨量计测值,而且能够准确反映雷达测得的降水分布形式.     

Calibration of a land-surface model using data from primary forest sites in Amazonia

Summary A land-surface model (MOSES) was tested against observed fluxes of heat, water vapour and carbon dioxide for two primary forest sites near Manaus, Brazil. Flux data from one site (called C14) were used to calibrate the model, and data from the other site (called K34) were used to validate the calibrated model. Long-term fluxes of water vapour at C14 and K34 simulated by the uncalibrated model were good, whereas modelled net ecosystem exchange (NEE) was poor. The uncalibrated model persistently underpredicted canopy conductance (g _c ) from mid-morning to mid-afternoon due to saturation of the response to solar radiation at low light levels. This in turn caused a poor simulation of the diurnal cycles of water vapour and carbon fluxes. Calibration of the stomatal conductance/photosynthesis sub-model of MOSES improved the simulated diurnal cycle of g _c and increased the diurnal maximum NEE, but at the expense of degrading long-term water vapour fluxes. Seasonality in observed canopy conductance due to soil moisture change was not captured by the model. Introducing realistic depth-dependent soil parameters decreased the amount of moisture available for transpiration at each depth and led to the model experiencing soil moisture limitation on canopy conductance during the dry season. However, this limitation had only a limited effect on the seasonality in modelled NEE.     

光流法及其在临近预报中的应用

基于多尺度IVAP(MIVAP,the Multi-scale Integrating Velocity Azimuth Process)方法反演风场进行临近外推的预报方法原理可以概括为：首先采用大尺度的分析单元通过IVAP(Integrating Velocity Azimuth Process)方法得到反演风场作为引导气流,表示雷暴的系统性移动,再通过小尺度的分析单元来反演风场,作为雷暴内部移动信息的表征,最后将不同尺度的风场进行合成得到外推的运动矢量场以进行临近预报。相比传统的多尺度TREC(MTREC,the Multi-scale Tracking Radar Echoes by Correlation)方法依赖于追踪回波过去的变化,需要相邻一个或者几个时次的雷达回波来跟踪雷达回波的移动,使用MIVAP方法外推预报只需要最新时刻的雷达反射率因子和径向速度。通过个例分析和统计分析,并且与MTREC方法进行对比试验,结果表明：MIVAP方法反演风场用于临近外推预报具有可行性,MIVAP方法具有实际应用的意义;MIVAP方法得到的外推运动矢量场方向要比MTREC方法方向总体整体偏右,且外推的平均速度要比MTREC方法快;MIVAP方法在30 min以内前几个时次的预报效果略低于MTREC方法,但是30 min以后的预报评分明显优于MTREC方法,MIVAP方法的检验评分整体要优于MTREC方法。

     

Analysis of quantitative precipitation forecasts using the Dynamic State Index

The German Weather Service (DWD) has two non-hydrostatic operational weather prediction models with different spatial resolution and precipitation parametrisations. The coarser COSMO-EU model has a spatial resolution of 7 km, whereas the higher-resolution COSMO-DE model has a gridspace of 2.8 km and explicitly resolves deep convection. To improve the numerical weather prediction (NWP) models it is necessary to understand precipitation processes. A central goal is the statistical evaluation of precipitation forecasts with dynamic parameters. Here, the Dynamic State Index (DSI) is used as a dynamic threshold parameter. The DSI theoretically describes the change of atmospheric flow fields as deviations from a stationary adiabatic solution of the primitive equations (Névir, 2004). For seasonal area means the DSI shows a remarkably high correlation with the precipitation forecasts provided by the COSMO-DE model. This is especially the case for the summer of 2007. The same analysis has been performed with the COSMO-EU forecast data and the results were compared with those from the COSMO-DE model. Moreover, an independent precipitation analysis, with a resolution corresponding to 7 km and 2.8 km, has been compared with respect to modelled precipitation and the DSI. In addition, correlations between the DSI and modelled as well as observed precipitation as a function of the forecast time for the different grid resolutions are also presented. The results show, that after 12 h, the correlation of the persistence forecast with the DSI reaches two thirds of the initial value. Thus, the DSI offers itself as a new dynamic forecast tool for precipitation events.     

Validation of BALTIMOS precipitation structures using BALTRAD radar data

The comparison of rain areas observed by the radar network BALTRAD and those produced by the model BALTIMOS shows that the model is able to reproduce the rain areas’ basic properties. These are power law distributed frequency distributions as well as lognormal rain area size distributions. The parameters governing these distributions are also matched by power laws evident for the observations and the BALTIMOS data. The distributions also provide evidence that the model produces too many large structures and too little small ones. However, the shape of these structures, represented by their fractal dimension, is well met. The perimeter of the model structures is only slightly smoother than that of the observed ones. The model is also able to reproduce the diurnal cycle of convective activity with emerging and decaying convective cells, though some deficiencies in the timing and the magnitude of the maximum occurrence of rain areas and peaks are evident. Additionally, it was found that the methods developed within this project are a valuable tool to validate BALTIMOS and potentially also other regional climate models.     

基于快速滚动更新的无缝隙定量降水预报模型

精细化定量降水网格预报是天气预报业务的重点和难点,基于西南区域智能数值网格模式预报系统(Southwest China WRF-based Intelligent Numeric Grid forecast System,SWC-WINGS)1 km× 1 km分辨率的小时降水预报,利用时间滞后和概率匹配方法开展融合订正试验,再利用中国气象局多源融合降水系统(CMA Multi-source Precipitation Analysis System,CMPAS)三源融合降水实况格点数据,对2022年7—8月四川地区的小时降水预报融合订正结果进行检验,并在四川盆地西部一次短时强降水天气过程中进行应用,结果表明：(1) 时间滞后集合降水预报相较于模式降水预报,存在小量级预报过度,大量级预报过于保守的问题;(2) 时间滞后结合概率匹配的降水预报融合订正方法有效提升了各量级降水预报的TS评分,尤其1~2 h预报时效提升显著,小时雨量超过0.1 mm、5 mm、10 mm和20 mm量级的TS评分平均提升率分别为7.2%、17.2%、28.3%和36.3%;(3) 一次短时强降水天气过程的应用结果表明,时间滞后结合概率匹配的融合订正方法对模式小时降水预报有较好的改进效果,尤其对大量级降水预报有较强的订正能力。

     

分类与集成方法在降雨预报中的应用

利用预报业务上常用的ECMWF (European Centre for Medium-Range Weather Forecasts)全球模式和华东区域模式(WARMS 2.0)的降水预报,建立了基于空间分布相似的降水集成方法,主要分为3个步骤：(1) 实时降水预报场空间尺度分离;(2)检索历史相似预报个例;(3)确定最优集成系数。在第(1)步尺度分离中采用高斯低通滤波方法,将降水场分解成连续性和分散性降水场;第(2)步相似个例检验利用了图像相似技术,根据连续性和分散性降水场综合寻找历史同期相似个例;在第(3)步集成系数确定中根据历史相似个例确定最优集成系数,并应用于最新实时预报中。通过结合图像识别、权重优化、建立历史样本数据库等方法,针对2018—2019年汛期(6—8月)检验结果表明：多模式集成产品的晴雨预报准确率相较于单模式而言有明显提高,且随时间变化表现较为稳定。在暴雨以上量级降水预报方面,集成产品的汛期整体TS评分高于单模式,Bias评分更接近1.0。通过CRA空间检验分析发现,集成产品既能一定程度上弥补全球模式在预报中尺度降水过程时强度偏弱的劣势,又纠正了区域模式在降水落区位置预报方面的偏差,进而实现了暴雨TS评分的提高。

     

Application of physical scaling towards downscaling climate model precipitation data

Physical scaling (SP) method downscales climate model data to local or regional scales taking into consideration physical characteristics of the area under analysis. In this study, multiple SP method based models are tested for their effectiveness towards downscaling North American regional reanalysis (NARR) daily precipitation data. Model performance is compared with two state-of-the-art downscaling methods: statistical downscaling model (SDSM) and generalized linear modeling (GLM). The downscaled precipitation is evaluated with reference to recorded precipitation at 57 gauging stations located within the study region. The spatial and temporal robustness of the downscaling methods is evaluated using seven precipitation based indices. Results indicate that SP method-based models perform best in downscaling precipitation followed by GLM, followed by the SDSM model. Best performing models are thereafter used to downscale future precipitations made by three global circulation models (GCMs) following two emission scenarios: representative concentration pathway (RCP) 2.6 and RCP 8.5 over the twenty-first century. The downscaled future precipitation projections indicate an increase in mean and maximum precipitation intensity as well as a decrease in the total number of dry days. Further an increase in the frequency of short (1-day), moderately long (2–4 day), and long (more than 5-day) precipitation events is projected.     

High-resolution precipitation data derived from dynamical downscaling using the WRF model for the Heihe River Basin,northwest China

Theoretical and Applied Climatology - The community of climate change impact assessments and adaptations research needs regional high-resolution (spatial) meteorological data. This study produced...     

Climatology of precipitation extremes in Estonia using the method of moving precipitation totals

A method of moving precipitation totals is described and applied for the analysis of precipitation extremes in Estonia. Numbers of extremely wet and extremely dry days and other indices of precipitation extremes were calculated using the daily precipitation data measured at 51 stations over Estonia during 1957–2009. Mean regularities of spatial and seasonal distribution were determined. Long-term changes were detected using Sen's method and Mann–Kendall test. The highest risk of heavy precipitation is in the regions of higher mean precipitation on the uplands and on the belt of higher precipitation in the western part of continental Estonia. Wet spells have their sharp maxima in July and August. The highest risk of droughts is observed in the coastal regions of West Estonia. In the coastal area, droughts appear mostly in the first half of summer, while in the eastern Estonia, they are usually observed during the second half of summer. Extreme precipitation events have become more frequent and intense. Statistically significant increasing trends were, first of all, found in the time series of winter extreme precipitation indices. In summer and autumn, trends existed in some indices, but in spring, there were no trends at all. There were no trends in time series of dryness indices in Estonia in 1957–2009.     

A short-range quantitative precipitation forecast algorithm using back-propagation neural network approach

A back-propagation neural network （BPNN） was used to establish relationships between the shortrange （0-3-h） rainfall and the predictors ranging from extrapolative forecasts of radar reflectivity, satelliteestimated cloud-top temperature, lightning strike rates, and Nested Grid Model （NGM） outputs. Quan- titative precipitation forecasts （QPF） and the probabilities of categorical precipitation were obtained. Results of the BPNN algorithm were compared to the results obtained from the multiple linear regression algorithm for an independent dataset from the 1999 warm season over the continental United States. A sample forecast was made over the southeastern United States. Results showed that the BPNN categorical rainfall forecasts agreed well with Stage Ⅲ observations in terms of the size and shape of the area of rainfall. The BPNN tended to over-forecast the spatial extent of heavier rainfall amounts, but the positioning of the areas with rainfall ≥25.4 mm was still generally accurate. It appeared that the BPNN and linear regression approaches produce forecasts of very similar quality, although in some respects BPNN slightly outperformed the regression.     

利用深度学习开展偏振雷达定量降水估测研究

利用2018—2020年经偏振升级改造后的广州S波段双偏振雷达（CINRAD/SAD）82892个体扫的0.5°仰角数据,以及雷达100 km探测范围内1109个雨量站共计538560个分钟雨量数据,分别构建了单参量、三参量雷达定量降水估测（QPE）深度学习网络架构（Z-Rnet、K_DP-Rnet、Pol-Rnet）,并以K_DP=0.5°/km为阈值分别训练得到大雨、小雨、总体等9个定量降水估测模型。在常用的均方误差作为损失函数的基础上,对不同降水强度采用不同权重提出了自定义损失函数,并利用比率偏差、相对偏差、均方差、平均绝对误差和平均相对误差作为评价指标对模型进行评估。通过对以积层混合云为主、以对流云为主和以层状云为主的3次降水过程的模型验证结果表明,利用深度学习训练的模型有较好的定量降水估测效果,区分雨强的小雨、大雨模型比不区分雨强的总体模型的效果要好。采用自定义损失函数模型效果更好,其均方差、平均绝对误差和平均相对误差分别较采用传统均方误差损失函数提升了8.62%、12.52%、16.34%。自定义损失函数中,采用Z_H-Z_DR-K_DP三参量网络架构训练得到的定量降水估测模型效果最好,其均方差、平均绝对误差和平均相对误差分别较采用Z_H的单参量Z-Rnet架构提升6.82%、8.43%、7.22%;较采用K_DP的单参量K_DP-Rnet架构提升12.33%、17.61%、17.26%。      

使用单部雷达探测资料进行甚短期降水预报

在统计分析合肥地区降水和强降水回波特征的基础上,提出并讨论了使用单部雷达探测资料进行测站降水甚短期预报的方法和能力。初步结果表明,该方法有一定的预报准确率,若综合考虑天气形势和卫星等其它探测资料,预报准确率会进一步提高。     

雷达识别冰雹云的综合指标方法

一、引言 及时地识别冰雹云,对于适时作业,搞好防雹具有极大的观实意义。测雨雷达识别冰雹云已在国内外广泛使用,但迄今为止还没有一个绝然的定量指标足以在雷达屏幕上把冰雹云和雷雨云明确地区别开来。辽宁省气象研究所和昭盟气象局从1973年起,利用“711”雷达在昭盟林西县进行了这方面的观测研究工作。五年来的观测经验证明,识别冰雹云需要综合地考虑一些因子,其中有定量的,如回波顶高度、强回波顶高度、负温区厚度和回波强度等;还有不定量的,如回波的外形结构特征及其演变规律等。本文利用多因子     

On the validation of the atmospheric model REMO with ISCCP data and precipitation measurements using simple statistics

Summary The regional atmospheric model REMO is used to study the energy and water exchange between surface and atmosphere over the Baltic Sea and its catchment area. As a prerequisite for such studies, the model has to be validated. A major part of such a validation is the comparison of simulation results with observational data. In this study the DX product of the International Cloud Climatology Project (ISCCP) and precipitation measurements from 7775 rain gauge stations within the model domain are used for comparisons with the simulated cloud cover and precipitation fields, respectively. The observations are available in this high spatiotemporal resolution for June 1993. To quantify the comparisons of means, variability, and patterns of the data fields simple statistics are used and the significance of the results is determined with resampling methods (Pool Permutation Procedure and Bootstrap-t). The conclusion is that simulated and observed means of the fields are not different at the 5% significance level. The determined variability of the fields is also in good agreement except the space variability in cloud cover. Time mean and anomaly patterns are in good coincidence in case of the comparisons of cloud cover fields, but in reduced coincidence in case of precipitation.With 9 Figures     

雨滴谱仪网数据在雷达定量降水估测中的应用

为了探讨利用雨滴谱数据验证雷达观测的回波强度的偏差和实时拟合反射率因子(Z)与降水强度(R)的关系(即Z-R关系)进行定量降水估测的可能性,以发生在江苏南部的三次大范围降水过程为例,首先分析雨滴谱数据、雷达数据和雨量计数据的一致性,然后利用雨滴谱仪网法和传统方法分别进行降水估测,并对比两种方法的降水估测效果。结果表明:雷达和雨滴谱仪观测的回波强度具有较好的一致性;采用雨量计数据和雨滴谱计算的平均降水强度存在一定差异,但其变化趋势基本一致。相对于层状云降水,两种方法对对流云降水估测的误差更大,对层状云降水的估测,雨滴谱仪网法略优于传统方法,但对对流云降水的估测,雨滴谱仪网法更具优势。总体上,雨滴谱仪网法估测的降水,其偏差和相对误差更小,估测值总量与雨量计观测雨量更接近,估测降水的效果更好。