Experimental real-time multi-model ensemble (MME) prediction of rainfall during monsoon 2008: Large-scale medium-range aspects

Realistic simulation/prediction of the Asian summer monsoon rainfall on various space–time scales is a challenging scientific task. Compared to mid-latitudes, a proportional skill improvement in the prediction of monsoon rainfall in the medium range has not happened in recent years. Global models and data assimilation techniques are being improved for monsoon/tropics. However, multi-model ensemble (MME) forecasting is gaining popularity, as it has the potential to provide more information for practical forecasting in terms of making a consensus forecast and handling model uncertainties. As major centers are exchanging model output in near real-time, MME is a viable inexpensive way of enhancing the forecasting skill and information content. During monsoon 2008, on an experimental basis, an MME forecasting of large-scale monsoon precipitation in the medium range was carried out in real-time at National Centre for Medium Range Weather Forecasting (NCMRWF), India. Simple ensemble mean (EMN) giving equal weight to member models, bias-corrected ensemble mean (BCEMn) and MME forecast, where different weights are given to member models, are the products of the algorithm tested here. In general, the aforementioned products from the multi-model ensemble forecast system have a higher skill than individual model forecasts. The skill score for the Indian domain and other sub-regions indicates that the BCEMn produces the best result, compared to EMN and MME. Giving weights to different models to obtain an MME product helps to improve individual member models only marginally. It is noted that for higher rainfall values, the skill of the global model rainfall forecast decreases rapidly beyond day-3, and hence for day-4 and day-5, the MME products could not bring much improvement over member models. However, up to day-3, the MME products were always better than individual member models.     

Deep learning for predicting the monsoon over the homogeneous regions of India

Indian monsoon varies in its nature over the geographical regions. Predicting the rainfall not just at the national level, but at the regional level is an important task. In this article, we used a deep neural network, namely, the stacked autoencoder to automatically identify climatic factors that are capable of predicting the rainfall over the homogeneous regions of India. An ensemble regression tree model is used for monsoon prediction using the identified climatic predictors. The proposed model provides forecast of the monsoon at a long lead time which supports the government to implement appropriate policies for the economic growth of the country. The monsoon of the central, north-east, north-west, and south-peninsular India regions are predicted with errors of 4.1%, 5.1%, 5.5%, and 6.4%, respectively. The identified predictors show high skill in predicting the regional monsoon having high variability. The proposed model is observed to be competitive with the state-of-the-art prediction models.     

Progress in Researches on Ensemble Forecasting of Extreme Weather Based on Numerical Models

Under the background of climate change, extreme weather events (e.g., heavy rainfall, heat wave, and cold damage) in China have been occurring more frequently with an increasing trend of induced meteorological disasters. Therefore, it is of great importance to carry out research on forecasting of extreme weather. This paper systematically reviewed the primary methodology of extreme weather forecast, current status in development of ensemble weather forecasting based on numerical models and their applications to forecast of extreme weather, as well as progress in approaches for correcting ensemble probabilistic forecast. Nowadays, the forecasting of extreme weather has been generally dominated by methodology using dynamical models. That is to say, the dynamical forecasting methods based on ensemble probabilistic forecast information have become prevailing in current operational extreme weather forecast worldwide. It can be clearly found that the current major directions of research and development in this field are the application of ensemble forecasts based on numerical models to forecasting of extreme weather, and its improvement through bias correction of ensemble probabilistic forecast. Based on a relatively comprehensive review in this paper, some suggestions with respect to development of extreme weather forecast in future were further given in terms of the issues of how to propose effective approaches on improving level of identification and forecasting of extreme events.     

利用WRF3D-Var同化多普勒雷达反演风场试验研究

为了将C波段雷达风场资料更好地应用于数值预报模式中,利用两步变分法反演多普勒雷达风场资料,并处理成标准的常规探空资料,以WRF模式及其三维变分同化系统为平台,针对2013年6月19日发生在天水的一次强暴雨过程进行同化雷达反演风的试验研究.试验结果表明:同化雷达反演风场后,对降水预报的改进能维持12h,尤其同化雷达反演风场后3~9h效果非常显著;0~3h作用不是很明显;9~12h预报具有一定的正作用.另外,循环同化比同化一次效果好,但并不是同化次数越多越好.因此,同化C波段雷达反演风场后,对降水预报具有一定的正作用.     

陕西省地质灾害气象预报预警方法研究

地质灾害气象预报预警方法是近年来地质灾害防范的热点。由于不确定因素较多,预报方法及预报精度上还有待提高。根椐我省实际情况,选取影响地质灾害发生的主要地质因素并设计了地质背景条件下各致灾因素概率模型;根椐地质灾害主要引发因素的降雨量和初步确定的降雨量临界值;选取BP神经网络模型算法进行数学运算,最终形成预报预警产品,分等级预报,在实际预报预警中取得较好的效果。     

数值天气预报检验方法研究进展

数值天气预报检验是改进及应用数值模式的重要环节。近年来,模式检验中的观念不断更新,适用于不同预报产品及不同用户需求的模式检验方法也不断涌现。首先简单回顾了以列联表为基础的传统的模式检验方法。其次重点总结了伴随高分辨率数值预报而出现的空间诊断检验技术,按照检验目的的不同,诊断方法可以归纳为:①基于滤波技术的分辨模式在不同时空尺度上预报能力的邻域法、尺度分离法;②利用位移偏差诊断模式预报位置、面积、方位、轴角等与观测差异的属性判别法、变形评估法。然后阐述了集合样本成员的概率分布函数(PDF)、集合预报与观测概率分布函数相似程度、事件发生的概率预报等集合预报检验方法。最后论述了空间诊断技术、集合预报检验方法的适用领域,并讨论了模式检验中存在的一些问题及未来的发展方向。     

数值模式的预报策略和方法研究进展

数值预报经历了半个多世纪的发展,已成为当前主要的客观预报工具。在模式和资料状况给定的情况下,预报效果的改善很大程度上依赖于所采用的预报策略和方法。为此,全面回顾了国内外基于数值模式的预报策略和方法研究进展,认为采取统计—动力相结合、从历史资料中提炼信息的预报策略是提高数值预报水平的可行之路。最后在总结前人工作基础上,着重介绍了动力相似预报策略和方法的相关研究,特别是实际预报中的试验情况。     

Application and Verification of Joint Probability Method in Potential Forecast for Severe Convective Weather in Anhui Province

In consideration of large uncertainties in severe convective weather forecast, ensemble forecasting is a dynamic method developed to quantitatively estimate forecast uncertainty. Based on ensemble output, joint probability is a post-processing method to delineate key areas where weather event may actually occur by taking account of the uncertainty of several important physical parameters. An investigation of the environments of little rainfall convection and strong rainfall convection from April to September (warm season) during 2009-2015 was presented using daily disastrous weather data, precipitation data of 80 stations in Anhui province and NCEP Final Analysis (FNL) data. Through ingredients-based forecasting methodology and statistical analysis,four convective parameters characterizing two types of convection were obtained, respectively, which were used to establish joint probability forecasting together with their corresponding thresholds. Using the ECMWF ensemble forecast and observations from April to September during 2016-2017, systematic verification mainly based on ROC and case study of different weather processes were conducted. The results demonstrate that joint probability method is capable of discriminating little rainfall convection and non-convection with comparable performance for different lead times, which is more favorable to identifying the occurrence of strong rainfall convection. The joint probability of little rainfall convection is a good indication for the occurrence of regional or local convection, but may produce some false alarms. The joint probability of strong rainfall convection is good at indicating regional concentrated short-term heavy precipitation as well as local heavy rainfall. There are also individual missing reports in this method, and in practice, 10% can be roughly used as joint probability threshold to achieve relative high TS score. Overall, ensemble-based joint probability method can provide practical short-term probabilistic guidance for severe convective weather.     

Ore Grade Prediction Using a Genetic Algorithm and Clustering Based Ensemble Neural Network Model

Accurate prediction of ore grade is essential for many basic mine operations, including mine planning and design, pit optimization, and ore grade control. Preference is given to the neural network over other interpolation techniques for ore grade estimation because of its ability to learn any linear or non-linear relationship between inputs and outputs. In many cases, ensembles of neural networks have been shown, both theoretically and empirically, to outperform a single network. The performance of an ensemble model largely depends on the accuracy and diversity of member networks. In this study, techniques of a genetic algorithm (GA) and k-means clustering are used for the ensemble neural network modeling of a lead–zinc deposit. Two types of ensemble neural network modeling are investigated, a resampling-based neural ensemble and a parameter-based neural ensemble. The k-means clustering is used for selecting diversified ensemble members. The GA is used for improving accuracy by calculating ensemble weights. Results are compared with average ensemble, weighted ensemble, best individual networks, and ordinary kriging models. It is observed that the developed method works fairly well for predicting zinc grades, but shows no significant improvement in predicting lead grades. It is also observed that, while a resampling-based neural ensemble model performs better than the parameter-based neural ensemble model for predicting lead grades, the parameter-based ensemble model performs better for predicting zinc grades.     

Ensemble Kalman filtering with shrinkage regression techniques

The classical ensemble Kalman filter (EnKF) is known to underestimate the prediction uncertainty. This can potentially lead to low forecast precision and an ensemble collapsing into a single realisation. In this paper, we present alternative EnKF updating schemes based on shrinkage methods known from multivariate linear regression. These methods reduce the effects caused by collinear ensemble members and have the same computational properties as the fastest EnKF algorithms previously suggested. In addition, the importance of model selection and validation for prediction purposes is investigated, and a model selection scheme based on cross-validation is introduced. The classical EnKF scheme is compared with the suggested procedures on two-toy examples and one synthetic reservoir case study. Significant improvements are seen, both in terms of forecast precision and prediction uncertainty estimates.     

GPM和TRMM卫星日降水数据在黄河源区的适用性分析

通过评估GPM计划三种日降水产品（IMERG-E、 IMERG-L和IMERG-F）和TRMM卫星、 两种日降水产品（TMPA 3B42和TMPA 3B42RT）在黄河源及其周边区域38个台站的适用性, 探究了五种产品探测精度和海拔高度及雨强的相关关系, 结果表明： 在与实测资料的一致性和偏差方面, GPM卫星产品要全面优于TMPA产品。在TRMM卫星产品中, 3B42产品明显优于3B42RT。五种产品的相关系数均表现出明显的从东南到西北递减的趋势, 均方根误差北部普遍低于南部。IMERG产品的探测率（POD）和探测成功率（CSI）都要普遍高于TMPA产品, 而误报率（FAR）则是TMPA 产品更低, 表现更好。五种产品均在个别台站出现了严重误报的情况, 这些台站主要分布在研究区的西北部。IMERG三种产品对于海拔高度的依赖程度具有很强的一致性, 而3B42RT产品对海拔高度几乎没有依赖。除3B42RT产品外, 其余四种产品的偏差均随雨强的增加而增大。在探测率方面, IMERG产品对小雨、 中雨和大雨的探测能力均优于TMPA产品。     

Probabilistic multi-model ensemble prediction of Indian summer monsoon rainfall using general circulation models: A non-parametric approach

Probabilistic prediction has the ability to convey the intrinsic uncertainty of forecast that helps the decision makers to manage the climate risk more efficiently than deterministic forecasts. In recent times, probabilistic predictions obtained from the products from General Circulation Models (GCMs) have gained considerable attention. The probabilistic forecast can be generated in parametric (assuming Gaussian distribution) as well as non-parametric (counting method) ways. The present study deals with the non-parametric approach that requires no assumption about the form of the forecast distribution for the prediction of Indian summer monsoon rainfall (ISMR) based on the hindcast run of seven general circulation models from 1982 to 2008. Probabilistic prediction from each of the GCM products has been generated by non-parametric methods for tercile categories (viz. below normal (BN), near-normal (NN), and above normal (AN)) and evaluation of their skill is assessed against observed data. Five different types of PMME schemes have been used for combining probabilities from each GCM to improve the forecast skill as compared to the individual GCMs. These schemes are different in nature of assigning the weights for combining probabilities. After a rigorous analysis through Rank Probability Skill Score (RPSS) and relative operating characteristic (ROC) curve, the superiority of PMME has been established over climatological probability. It is also found that, the performances of PMME1 and PMME3 are better than all the other methods whereas PMME3 has showed more improvement over PMME1.     

Displacement prediction model of landslide based on a modified ensemble empirical mode decomposition and extreme learning machine

In this paper, an M–EEMD–ELM model (modified ensemble empirical mode decomposition (EEMD)-based extreme learning machine (ELM) ensemble learning paradigm) is proposed for landslide displacement prediction. The nonlinear original surface displacement deformation monitoring time series of landslide is first decomposed into a limited number of intrinsic mode functions (IMFs) and one residual series using EEMD technique for a deep insight into the data structure. Then, these sub-series except the high frequency are forecasted, respectively, by establishing appropriate ELM models. At last, the prediction results of the modeled IMFs and residual series are summed to formulate an ensemble forecast for the original landslide displacement series. A case study of Baishuihe landslide in the Three Gorges reservoir area of China is presented to illustrate the capability and merit of our model. Empirical results reveal that the prediction using M–EEMD–ELM model is consistently better than basic artificial neural networks (ANNs) and unmodified EEMD–ELM in terms of the same measurements.     

Spatial and temporal monthly precipitation forecasting using wavelet transform and neural networks,Qara-Qum catchment,Iran

This paper aims to provide a spatial and temporal analysis to prediction of monthly precipitation data which are measured at irregularly spaced synoptic stations at discrete time points. In the present study, the rainfall data were used which were observed at four stations over the Qara-Qum catchment, located in the northeast of Iran. Several models can be used to spatially and temporally predict the precipitation data. For temporal analysis, the wavelet transform with artificial neural network (WTANN) framework combines with the wavelet transform, and an artificial neural network (ANN) is used to analyze the nonstationary precipitation time-series. The time series of dew point, temperature, and wind speed are also considered as ancillary variables in temporal prediction. Furthermore, an artificial neural network model was used for comparing the results of the WTANN model. Therefore, four models were developed, including WTANN and ANN with and without ancillary data. Several statistical methods were used for comparing the results of the temporal analysis. It was evident that at three of the four stations, the WTANN models were more effective than the ANN models, and only at one station, the ANN model with ancillary data had better performance than the WTANN model without ancillary data. The values of correlation coefficient and RMSE for WTANN model with ancillary data for the validation period at Mashhad station which showed the best results were equal to 0.787 and 13.525 mm, respectively. Finally, an artificial neural network model was used as an alternative interpolating technique for spatial analysis.     

典型降水预报ANN多指标优选——以太湖湖西区为例

为提高多种典型人工神经网络应用于降水预报的精度与稳定性并做出优选,对太湖流域湖西区丹徒、丹阳、金坛、溧阳、宜兴5站的年降水量时间序列建立基于组成成分分析的人工神经网络模型,并通过平均相对误差、平均绝对误差、均方根误差及合格率4项评价指标对比分析预报效果。该模型采用Mann-Kendall法、秩和检验法、谱分析法进行组成成分分析;建立BP网络、小波神经网络、RBF网络、GRNN网络及Elman网络模拟并预测随机成分,与确定性成分叠加得年降水量预报结果。在湖西区的研究结果表明,基于组成成分分析的人工神经网络模型的拟合及预测精度高于原始人工神经网络和线性自回归模型,GRNN网络的预测精度与稳定性高于其他4类神经网络。     

北京山区基于历史资料的泥石流临界雨量研究

通过对北京山区历史泥石流降雨资料的分析,根据有雨量站和无雨量站的泥石流易发区的资料情况,分别建立了不同前期雨量的预报模型,并根据预报模型初步建立了不同前期雨量前提条件下的黄,橙,红三色预报模式。     

An Overview of Ocean Predictability and Ocean Ensemble Forecast

Ocean is a highly complex and nonlinear dynamical system. The inevitable errors in both data and numerical models lead to uncertainties in ocean numerical prediction. By understanding features and properties in the ocean on multiple scales, it is important to quantify and estimate the predictability of the ocean, and analyze the reasons and mechanism of error growth. The efforts focus on investigating the method to reduce the uncertainties and errors in forecasting and increase the time limit of ocean predictability. The advances will result in improved marine forecasting models and forecasting skill. Understanding limitations and identifying the research needed to increase accuracy will lead to fundamental progress in ocean forecast, which is of great significance. The present study described and illustrated the mechanics and computations involved in modeling and predicting uncertainties for ocean prediction and its modern applications. Firstly, it discussed the fundamental concept and classification of the ocean predictability. The research status of ocean predictability is introduced including the dynamics methodologies and the ocean ensemble prediction. Three of the dynamical computational methodologies including the singular vector, Lyapunov exponent and bred vector method were introduced. Three ocean ensemble prediction methods including initial condition ensemble, multi-model ensemble and atmospheric forcing ensemble were described and illustrated. Finally, this paper gave a future prospective of ocean predictability and its application.     

Assessment of climate change impacts on rainfall using large scale climate variables and downscaling models – A case study

Many of the applied techniques in water resources management can be directly or indirectly influenced by hydro-climatology predictions. In recent decades, utilizing the large scale climate variables as predictors of hydrological phenomena and downscaling numerical weather ensemble forecasts has revolutionized the long-lead predictions. In this study, two types of rainfall prediction models are developed to predict the rainfall of the Zayandehrood dam basin located in the central part of Iran. The first seasonal model is based on large scale climate signals data around the world. In order to determine the inputs of the seasonal rainfall prediction model, the correlation coefficient analysis and the new Gamma Test (GT) method are utilized. Comparison of modelling results shows that the Gamma test method improves the Nash–Sutcliffe efficiency coefficient of modelling performance as 8% and 10% for dry and wet seasons, respectively. In this study, Support Vector Machine (SVM) model for predicting rainfall in the region has been used and its results are compared with the benchmark models such as K-nearest neighbours (KNN) and Artificial Neural Network (ANN). The results show better performance of the SVM model at testing stage. In the second model, statistical downscaling model (SDSM) as a popular downscaling tool has been used. In this model, using the outputs from GCM, the rainfall of Zayandehrood dam is projected under two climate change scenarios. Most effective variables have been identified among 26 predictor variables. Comparison of the results of the two models shows that the developed SVM model has lesser errors in monthly rainfall estimation. The results show that the rainfall in the future wet periods are more than historical values and it is lower than historical values in the dry periods. The highest monthly uncertainty of future rainfall occurs in March and the lowest in July.     

Skills of different mesoscale models over Indian region during monsoon season: Forecast errors

Performance of four mesoscale models namely, the MM5, ETA, RSM and WRF, run at NCMRWF for short range weather forecasting has been examined during monsoon-2006. Evaluation is carried out based upon comparisons between observations and day-1 and day-3 forecasts of wind, temperature, specific humidity, geopotential height, rainfall, systematic errors, root mean square errors and specific events like the monsoon depressions.It is very difficult to address the question of which model performs best over the Indian region? An honest answer is ‘none’. Perhaps an ensemble approach would be the best. However, if we must make a final verdict, it can be stated that in general, (i) the WRF is able to produce best All India rainfall prediction compared to observations in the day-1 forecast and, the MM5 is able to produce best All India rainfall forecasts in day-3, but ETA and RSM are able to depict the best distribution of rainfall maxima along the west coast of India, (ii) the MM5 is able to produce least RMSE of wind and geopotential fields at most of the time, and (iii) the RSM is able to produce least errors in the day-1 forecasts of the tracks, while the ETA model produces least errors in the day-3 forecasts.     

Fast linearized forecasts for subsurface flow data assimilation with ensemble Kalman filter

Ensemble methods present a practical framework for parameter estimation, performance prediction, and uncertainty quantification in subsurface flow and transport modeling. In particular, the ensemble Kalman filter (EnKF) has received significant attention for its promising performance in calibrating heterogeneous subsurface flow models. Since an ensemble of model realizations is used to compute the statistical moments needed to perform the EnKF updates, large ensemble sizes are needed to provide accurate updates and uncertainty assessment. However, for realistic problems that involve large-scale models with computationally demanding flow simulation runs, the EnKF implementation is limited to small-sized ensembles. As a result, spurious numerical correlations can develop and lead to inaccurate EnKF updates, which tend to underestimate or even eliminate the ensemble spread. Ad hoc practical remedies, such as localization, local analysis, and covariance inflation schemes, have been developed and applied to reduce the effect of sampling errors due to small ensemble sizes. In this paper, a fast linear approximate forecast method is proposed as an alternative approach to enable the use of large ensemble sizes in operational settings to obtain more improved sample statistics and EnKF updates. The proposed method first clusters a large number of initial geologic model realizations into a small number of groups. A representative member from each group is used to run a full forward flow simulation. The flow predictions for the remaining realizations in each group are approximated by a linearization around the full simulation results of the representative model (centroid) of the respective cluster. The linearization can be performed using either adjoint-based or ensemble-based gradients. Results from several numerical experiments with two-phase and three-phase flow systems in this paper suggest that the proposed method can be applied to improve the EnKF performance in large-scale problems where the number of full simulation is constrained.