Vol.29, NO.1, 2018

The Architecture Design of MICAPS4 Server System
Wang Ruotong, Wang Jianmin, Huang Xiangdong, Dong Yifeng, Long Mingsheng
2018, 29(1): 1-12. DOI: 10.11898/1001-7313.20180101
Meteorological data are typical non-structure data, which reach dozens of TBs per day. Data pre-processing, data storage and data access based on RDBMS and file system become the bottleneck of MICAPS3. To fulfill MICAPS4 users' need of fast, in-time query of meteorological real-time data, according to the multi-dimension model and the user query behavior of meteorological data, using non-relational key-value DDBMS, a high performance massive meteorological data storage system and a stable 7×24 distributed data pre-processing system is designed and established. MICAPS4 uses a client/server system architecture, and high-performance server cluster system is the critical component of MICAPS4. Using distributed key-value data model and P2P infrastructure, MICAPS4 server system distributes all real-time data which arrive at a very high speed to multiple servers through an automatic load balance algorithm, and all data are stored in memory initially and persistent to hard disk periodically, which can not only reduce the disk I/O operating times, but also guarantee the reduction of writing pressure accompanying the high load of reading pressure. To enhance the data and system reliability, distributed system architecture and multiple data replica are used, which also improves the throughput capacity of the system. According to statistic results gained from product environment, the performance of MICAPS4 server system improves 100 times more than MICAPS3. MICAPS4 server system transits all meteorological real-time data storage from file system to database, from centralized system to distributed system. The system becomes the core production system of China Meteorological Administration in 2015 and is popularized nationwide. Under the condition of massive meteorological data and concurrent access of many users, it shows high stability and excellent read-write performance, and it is also highly scalable and maintenance friendly. MICAPS4 high performance server system includes 5 sub-systems including distributed storage system, distributed pre-processing system, station data polling system, data query server and monitoring probe. The distributed storage system provides high performance data accessing services of meteorological real-time data in both random and sequence mode, the distributed pre-processing system implements the stream computing function of massive meteorological real-time data by adopting the peer to peer distributed system infrastructure, the station data polling system implements the heterogeneous station observation replica data synchronization function over different systems, the data query server implements MICAPS4 client real-time computing function by means of the multi-threading server technology, and the monitoring probe is deployed in each server node and reports host health messages periodically. The overall design of MICAPS4 server system is depicted, and the motivation, core technologies and the design of each sub-system are also introduced.
Design and Implementation of Intelligent Grid Forecasting Platform Based on MICAPS4
He Yanan, Gao Song, Xue Feng, Zhao Shengrong, Liu Ming, Hu Hao, Wei Tao
2018, 29(1): 13-24. DOI: 10.11898/1001-7313.20180102
With the rapid development of modern weather prediction in China, large amounts of high-resolution data are widely used. An efficient and convenient platform for data displaying, analyzing and forecasting is urgently needed. In response to the requirement of business, an intelligent grid forecasting platform is designed and implemented based on MICAPS4 (Meteorological Information Comprehensive Analysis Processing System Version 4), which is used by the provincial meteorological departments to produce and distribute meteorological grid forecasting products. The application background, requirement analysis, framework design and main functions implementation are discussed in details. In addition, some key technologies involved in the platform realization are also described.The MVVM (model-view-viewmodel) design mode is adopted in the platform to separate the business logic from the view. And the coupling degree between modules is reduced by dividing each sub-function module, which has good scalability. The supporting data environment of the platform mainly includes meteorological service network/Internet, CIMISS (China Integrated Meteorological Information Service System), as well as other local data sources. The display and analysis program of high-resolution grid forecast data is realized, which includes 17 weather elements such as precipitation, temperature, wind, relative humidity, cloud cover and disastrous weather. And several intelligent forecasting tools based on contours, grids and key points are developed, which integrates objective forecast methods such as the precipitation time consistency algorithm and the correction of timing temperature using 24 h high and low temperature extreme value. Some element consistency algorithms are developed, such as precipitation, temperature, humidity and so on, which helps forecasters to improve work efficiency and ensure consistency among products. A graphical configuration management interface is provided, which facilitates user localization application.Since July 2016, the platform has been put into operation in most of the provincial meteorological departments in China, which provides an important support for the national intelligent grid forecasting business. Ever since, the platform is further improved by adding new functions and fixing bugs based on user feedbacks.In the future, intelligent multi-source objective prediction products recommending module will be developed, which is based on machine learning, inspection and evaluation and other technical methods. An intelligent collaborative engine of elements will be designed and realized in order to achieve more diverse forecast products. And some of the classic objective methods in weather forecasting will be integrated into the platform. On this basis, the visual modeling tool will be developed, which provides a graphical modeling approach to model forecasting experience, so that experience and intelligent methods of forecasters can achieve a better combination.
Design and Implementation of SWAN2.0 Platform
Han Feng, Wo Weifeng
2018, 29(1): 25-34. DOI: 10.11898/1001-7313.20180103
Severe Weather Automatic Nowcasting System 2.0(SWAN2.0) is a short-term nowcasting operational platform of CMA, providing nowcasting products and an early warning product generation tool. SWAN2.0 includes three types of meteorological products. Observation products, mainly composed of radar puzzles and automatic weather station(AWS) observations. Alarm products, including AWS elements alarms and radar echo alarms. Nowcasting products, providing 0-1 h radar echo forecast by COTREC movement vector and the tracking and forecasting of convection storm by SCIT (Storm Cell Identification and Tracking) or TITAN(Thunderstorm Identification, Tracking, Analysis and Nowcasting). SWAN2.0 is based on MICAPS4(Meteorological Information Comprehensive Analysis and Processing System Version 4) development framework, using C/S architecture. The server of SWAN2.0 is a scheduling platform of meteorological algorithm, which is deployed at the provincial meteorological administration, in charge of collecting data, running algorithm, and generating SWAN products. The client of SWAN2.0 is a complete working platform for weather forecasters deployed in national, provincial, and municipal meteorological observatories, which are used to display SWAN products, make analysis and produce weather forecast products. SWAN2.0 adopts new nowcasting technologies, such as three-dimensional variation assimilation retrieval of wind field, QPE(quantity precipitation estimation) by rain cluster, hail identification and meso-scale numerical model application, supporting weather forecasters to extend from traditional short-term weather forecasts and services to short-range and nowcasting forecasts of classified strong convective weather.SWAN2.0 integrates computer technology and forecasting technology to solve short-term forecasting problems. It uses the message queue to decouple business modules to enhance the flexibility and scalability of the platform, and can generate early warning produces automatically from alarm products. The hierarchical structure is adopted to optimize the design of the alarm module, and the alarm module efficiency is improved with pipeline filter model and asynchronous technology.In addition, SWAN2.0 adds two common data models, grid data model and feature data model, creating easy access to local products.In short, SWAN2.0 is not only a operational platform for forecaster but also a set of open data platform and development environment. It provides data services of real-time radar, automatic station and basic short-term nowcasting data, open operating environment and display terminal for the station, and provides support for station localization algorithm development.SWAN2.0 is released in July 2016, and popularized in nationwide. It provides an important foundation and reference for routine nowcasting operation.
Development and Application of Typhoon and Marine Meteorological Integrated Operational System
Cao Li, Gao Song, He Yanan, Zhao Wei, Qian Qifeng
2018, 29(1): 35-44. DOI: 10.11898/1001-7313.20180104
Typhoon and Marine Meteorological Integrated Operational System, designed for typhoon and marine meteorological warning and forecasting operations, is an integrated platform based on MICAPS4 (Meteorological Information Comprehensive Analysis and Processing System Version 4) framework, providing functions of information retrieval, displaying of typhoon track, marine meteorological observations and numerical forecast products, as well as service product production, distribution and other functions.Typhoon and Marine Meteorological Integrated Operational System consists of three layers, namely, data layer, application component layer and operational service layer. The data layer refers to the database, file system and corresponding processing and management system of the application server. It mainly includes the management and maintenance of the typhoon real-time database, the marine meteorological product database, the meteorological information base sharing information. It collects and transforms the typhoon marine meteorological data, and preprocesses the typhoon marine meteorological model outputs for fine forecasts. This layer provides data services for application component layers and operational service layers. The application component layer contains the basic components and functional components. The basic components mainly refer to components based on MICAPS4 kernel framework, including the underlying services such as graphics engine, basic management, plug-in management, service management and GIS components. Function components mainly include data analysis, streamline analysis, contour analysis, interactive tools, product production, product release and other basic functions. These components provide support for the operational service layer through the interface service. The operational service layer covers all functions of the service, which are used to display and receive data and provide an interactive interface. It integrates typhoon, marine meteorological data retrieval, data display, fine grid editing, typhoon track production, product production, product release and other applications, through the interactive operation, to achieve operational early warning and forecasting function.Typhoon and Marine Meteorological Integrated Operational System has been put into use as the main platform of typhoon and marine meteorological forecasting and warning service in China National Meteorological Center. With functions of comprehensive retrieval and display of meteorological data, fine grid editing, forecast and service product making, product publication, and other basic operational functions, the system significantly enhances the fine forecasting ability of typhoon and marine meteorological. It also shows good operational application ability and development prospects.
Design and Implementation of MICAPS4 Web Platform
Hu Zhengguang, Gao Song, Xue Feng, Yu Lianqing
2018, 29(1): 45-56. DOI: 10.11898/1001-7313.20180105
In recent years, Web-based meteorological application platform plays a key role in meteorological data sharing, weather forecast cooperation and early-warning, weather service and meteorological decision-making application. Many meteorological Web service systems are developed based on flash plugins or commercial software such as ArcGIS. For instance, French Meteorological Department develops Synergie-Next, German Meteorological Department develops NINJO, and ECMWF develops MetView adopting WebGIS to integrate the WMS and WMF map services. However, there are still some limitations, for example, these applications cover only limited kinds of meteorological data, and it is expensive and inconvenient to develop WebGIS applications based on third-part plugins or commercial GIS software. Besides, some systems are platform-dependent and not robust when run on different operation systems. Finally, some Web-based applications don't support meteorological data analysis and interoperation well.To solve these problems, MICAPS4 Web Platform is developed by National Meteorological Center(NMC) using JAVA and the Brower/Server mode, and it integrates various models including the VECTOR Model, RASTORV Model, GIRD Model, which can store, process different kinds of meteorological data efficiently. It can be used to publish kinds of heterogeneous meteorological data sources through Web efficiently, and with the help of integrated analysis algorithms and models, this platform could be used for developers to develop kinds of meteorological applications systems. Meanwhile, some modern key technologies are used in this platform, such as drawing based on HTML5, and meteorological data real-time distributed process computation with STORM. MICAPS4 Web Platform provides efficient and common server-side and browser-side interface API, making it a convenient meteorological Web platform which could be used by developers at different levels of meteorological departments.MICAPS4 Web Platform and other applications developed by national or some provincial meteorological observatories with its SDK run stably in real-time weather operation, which illustrates strong practicability and expansibility in massive data Web publishing, Web client rendering, forecast analysis and interoperation, and data monitoring. MICAPS4 Web Platform provides efficient server-side and browser-side SDK API, through which national developers already develop kinds of metrological forecasting and early-warning information platforms, and meanwhile, local developers can also develop local meteorological Web information platforms quickly. There are still some problems to be solved further, such as researches about micro-services and simplifying these SDK APIs.
The Fog/Haze Medium-range Forecast Experiments Based on Dynamic Statistic Method
Zhang Ziyin, Zhao Xiujuan, Xiong Yajun, Ma Xiaohui
2018, 29(1): 57-69. DOI: 10.11898/1001-7313.20180106
Beijing and eastern China have frequently suffered from severe fog/haze days in recent years, which are characterized by high particle mass concentration and low visibility. Severe haze/fog pollution, especially the persisted fog/haze days (i.e., in January 2013 and November, December of 2015) greatly threaten human health and traffic safety. These phenomena stimulate great interest in studying the fog/haze pollutions in Beijing or even eastern China. The fog/haze pollution is in general attributed to two aspects:Pollutants emission to the lower atmosphere from fossil fuel combustion, construction and others, and unfavorable meteorological diffusion conditions. Air quality or the occurrence of fog/haze pollution are strongly influenced by meteorology. Meteorological factors not only have essential impacts on the accumulation or diffusion, spread and regional transport of air pollutants, but also have important impacts on the formation of secondary aerosol which are generated by the complicated physical and chemical reactions. Particularly, weather conditions play an essential role in the daily variability of air pollutant concentrations.Based on the dynamic statistic forecasting method and the high-resolution weather forecast fields derived from European Centre for Medium-Range Weather Forecasts (ECMWF), the fog/haze medium-range forecast system is designed to provide objective and quantitative PM2.5 and visibility forecasts for cities in Beijing-Tianjin-Hebei and its adjacent regions by predicting 1 to 10 days in advance. A forecasting experiment is performed during the period from 1 October 2015 to 10 November 2016. Results show that the predicted PM2.5 concentrations and visibilities based on the method for 14 cities (Beijing, Tianjin, Shijiazhuang and others) and different leading times (namely 1 to 10 days in advance) are well consistent with the observed. All correlation coefficients between them are significant at 0.01 level. And most of the reduction errors (RE) between them are larger than 0.2. Most of TS values are confined in 0.1 to 0.3, and the mean of all TS values are close to 0.2 for visibility, PM2.5 grades and fog/haze phenomena. Moreover, several case analyses suggest that the method can predict the change trends of the continuous fog/haze process about 5-6 days in advance. Generally, the method can approximately predict the hourly variability of the PM2.5 concentration and visibility and the change trends of the process of fog/haze and heavy pollution in Beijing-Tianjin-Hebei and its adjacent regions on the medium-range time scale. The high reliability and stability of the forecasting test suggest that the objective and quantitative predictions produced by the method can be used with high reference value for the medium-range forecast of fog/haze and air quality in Beijing and surrounding cities.
Contrastive Analysis of Two Intense Typhoon-tornado Cases with Synoptic and Doppler Weather Radar Data in Guangdong
Huang Xianxiang, Yu Xiaoding, Yan Lijun, Li Zhaoming, Li Cailing
2018, 29(1): 70-83. DOI: 10.11898/1001-7313.20180107
Conventional observations, Doppler weather radar and NCEP/NCAR reanalysis data are used to analyze two strong tornado events originated in the outside-region of typhoons on 4 October 2015(EF3) and 4 August 2006(EF2) contrastively. Results show that two strong typhoon-tornado events both occur in the northeast quadrant of landfalling typhoons, with some similar environmental conditions including low-level convergence, upper-level divergence and superimposition of strong southeast jet at mid and low level over the Pearl River Delta. The difference of circumstances is that two typhoons are in different weakening phase. The former typhoon has just landed and is more organized and stronger than the latter. Environmental parameters are shown to be relatively moderate convective available potential energy, low convection inhibition, low condensation uplift height, strong deep (0-6 km) and low level (0-1 km) vertical wind shear and large storm relative helicity (SRH). Storm relative helicity is a good indicator for the occurrence of supercell or mesocyclone. The larger the SRH is, the more likely a supercell or mesocyclone may form. Combining the northeastern quadrant of typhoon with the high SRH area, the area where typhoon tornadoes may occur could be determined to a certain extent. The two tornado storms are mini-supercells, and the radar base reflectivity factors of two tornadoes are similar to features of classical supercell such as the low-level warm-humid inflow gaps and hook echoes. The former (tornado parent storm on 4 October 2015) has stronger echo and more apparent hook echo features. Strong mesocyclones and tornado vortex signature (TVS) can be observed on radar speed chart in both tornado events, and mesocyclones form at mid-low level firstly, then develop to the lower level, resulting in tornadoes finally. TVS is observed either synchronously with the tornado touchdown or a volume scanning ahead. Vertical vorticity of TVS in the center of low-level mesocyclones is strong, and the bottom and top heights are very low in the mesocyclone and TVS. The difference of bottom/top height of the mesocyclone and TVS between two cases is that, the former presents an abruptly-drop phenomenon whereas the latter (tornado parent storm on 4 August 2006) maintains at low level before and after the tornado touchdown. Before and after tornadoes touchdown, the strongest wind shear of storms both increase sharply, but the wind shear in TVS is larger, which is about twice of the latter.
Identification and Correction of the Bright Band Using a C-band Dual Polarization Weather Radar
Cao Yang, Chen Hongbin, Su Debin
2018, 29(1): 84-96. DOI: 10.11898/1001-7313.20180108
The bright band is a layer of enhanced reflectivity due to melting of aggregated snow and ice crystals. The occurrence of a bright band causes significant overestimation in radar-based quantitative precipitation estimation (QPE). The bright band signature can be normally identified from vertical profiles of reflectivity (VPRs) of stratiform precipitation echoes and the freezing level height which is derived from radiosonde data. The VPRs correction is desirable to mitigate the bright band contamination and reduce the overestimation of the radar-based QPE. However, a well-defined bright band bottom, which is critical for the correction of bright band, is sometimes not found in VPRs. Fortunately, polarimetric variables, especially the correlation coefficient, can provide a much better depiction of vertical bright band structure than reflectivity.The volume scanning data of a C-band dual polarization radar from Beijing Meteorological Bureau, radiosonde data and measured rainfall data from ground rain gauge stations are used to test the methodology of the bright band identification and correction. Three bright band correction schemes including mean vertical profile of reflectivity (MVPR), apparent vertical profile of reflectivity (AVPR) and apparent vertical profile of correlation coefficient (AVPCC), which are derived from stratiform precipitation echoes, are applied to the reflectivity field in the given tilt, and radar-based QPEs are derived from the corrected reflectivity field based on traditional Z-R relations. Results indicate that the bright band top, peak and bottom can be easily identified from the volume scanning MVPR and the freezing level height, and most of bright band depths are between 0.8 km and 1.5 km. The AVPR and AVPCC schemes are shown to be more effective in mitigating the bright band contamination and reducing the overestimation of radar-derived QPE associated with the bright band than the MVPR correction. Corrected reflectivity fields are physically continuous in distribution, and the corrected radar-derived QPEs are close to the measured value of ground rain gauge stations.
Impacts of Assimilating Wind Profiler Radar Observations on Precipitation Prediction in Zhejiang Province
Yu Zhenshou, Ji Chunxiao, Yang Chen, Li Yuejun
2018, 29(1): 97-110. DOI: 10.11898/1001-7313.20180109
Wind profiler radar (WPR) is a new type of wind measuring radar, which has advantages of high spatial resolution, continuity and good instantaneity. With the increase of wind profile radar year by year, it is meaningful to apply this kind of wind field observations to the numerical model to improve the model prediction ability. The meso-scale numerical prediction model WRF and the assimilation system ADAS developed by Center for Analysis and Prediction of Storms, University of Oklahoma, is used to study effects of assimilating observations of 35 wind profiler radars in eastern China on precipitation prediction over Zhejiang. Prior to assimilation, 1 h average sampling product data are subjected to climate extreme inspection, consistency check and vertical thinning for quality control. A spring rainstorm process on 16-17 May 2014 is selected as an example to evaluate effects of WPR data assimilation on the quality of precipitation forecast in detail. And effects of WPR data are also verified by batch experiments starting from 0000 UTC and 1200 UTC during the whole June of 2015. Results show that the model precipitation TS and ETS scores are improved, especially for heavy rainfalls. At the same time, the false alarm ratio (FAR) and frequency of misses (FOM) for heavy and torrential rain decrease after WPR data assimilation, but the FAR of moderate rain increase. The case study shows that WPR data assimilation can adjust the initial field of low layer wind field, increase small scale weather information, and improve the horizontal wind prediction on the whole layers. For 12 h wind forecast field, the result of assimilation of WPR is obviously better than that without the assimilation. In addition, the improvement of the zonal wind is more obvious than that of the meridional wind after WPR data assimilation. The case study shows that 850 hPa wind speed is enhanced by 20%-30%, water vapor flux is increased by 30%-50%, and the atmospheric instability in the rainstorm area and its upstream region is also enhanced after WPR data assimilation. As a result, TS of light rain and heavy rain is increased by 0.06-0.07, and FAR and FOM of rainstorm is reduced by 0.04-0.05. Although the assimilation of wind profiler data can improve the precipitation prediction quality, there are still some problems, such as an unexplained overestimation of regional average precipitation, which needs further investigation.
A Method for Summer Maize Phenology Monitoring by MODIS Data
Li Ying, Chen Huailiang, Li Yaohui, Wang Xiuping, Zhang Fangmin
2018, 29(1): 111-119. DOI: 10.11898/1001-7313.20180110
Crop phenology period is an important feature of the agricultural eco-system. It is important to investigate the crop phenology period in large area for precision crop management and yield forecast by remote sensing technical. However, there are still some limitations in this approach, such as the investigation precision is restricted by the investigation area, and different types of crop growth curves don't match with the crop phenology period very well.Therefore, data from moderate-resolution imaging spectroradiometer (MODIS), and summer maize growth stages by field observations from 23 agricultural meteorological stations in Henan Province are adopted to improve the identification efficiency and accuracy. Normalized difference vegetation index (NDVI) growth curve with the time resolution of 1 d is reconstructed by denoising, smooth processing and logistic curve fitting. Crop growth feature points on the reconstructed growth curve are extracted by using dynamic threshold method and curvature extremum method. The optimum matching relationship between feature points and maize growth stages is constructed by based on the feature points, their occurrence date, and observed dates of growth stages. By matching with the investigated growth stage dates in 2013-2014, values of dynamic threshold 1 is chosen to extract the 7-leaf stage, with the root mean square error (RMSE) of 5.4 d. Values of minimum curvature is chosen to extract the jointing stage, with the RMSE of 6.4 d. Values of dynamic threshold 2 is chosen to extract the tasseling stage, with the RMSE of 6.0 d. By validation with the investigated growth stage dates in 2015, RMSE with the selected 3 key growth stages are all less than 6 d. The accuracy is higher than the earlier proposed methods to extract maize growth stages by using MODIS or other similar lower/medium spatial resolution remote sensing data. Map of summer maize critical phenology in Henan Province demonstrates that most of summer maize in the research areas enter 7-leaf stage, jointing stage and tasseling stage in June 25 to 27 June, 10 July to 17 July and 27 July to 2 August, respectively. Numbers of pixels, which enter 7-leaf stage, jointing stage and tasseling stage on the above-mentioned dates, account for 45.6%, 66.8% and 71% of the total, respectively. Maize growth stages obtained by the method proposed by this research can be used in crop management and grain yield forecast.
WebGIS-based Meteorological Service System and Its Key Technology
Lü Zhongliang, Bai Xinping, Xue Feng
2018, 29(1): 120-128. DOI: 10.11898/1001-7313.20180111
In response to the urgent need of online meteorological service products in business, the construction of online production system of meteorological service products is carried out based on service integration. Details of the key technology is discussed, including dynamic service registration technology, product and service standardization, product service pipelined realization, the application of comprehensive regular operation rules in dynamic updating of parameters, and multi-level tasks scheduling. The system consists of server and client sides, and the server takes the duty of service interface development and release, while the client can be used for the management and operation of the task scheduling by all levels of the user.The WebGIS architecture is used in the weather service product online production system. The server is responsible for the development and management of the service interface, solving the problem of product production efficiency and standardize through distributed computing. It also packs the research and development of service component development module, service model creation module, service template development module, service configuration module, and service package module. The client together with the server side solves problems of system resource requirements and maintenance, also, the client achieves the function of task query, task management, operation monitoring, data statistics, user management, and parameter configuration.The construction of service-based system is discussed in detail too. The whole system is composed of services, using a dynamic, distributed and service-oriented system framework. Components are the core of the whole architecture, and they use and provide a variety of services that can be found. The service is the consistent window for the entire system, the product is a service, the interface is a service, and even the whole system is a service. Users do not need to care about the use of the system hardware and software deployment, product specific production steps and data sources. Users only need to analyze requirements and use the service interface on-demand production or extract results of the interface.Dynamic expansion of services and flexible configuration are realized through the service dynamic registration technology. Business products, overall norms, upgrades, efficiency is guaranteed by the key technology of product and service standardization and product service pipelined realization. The application of comprehensive regular operation rules improves the flexibility of the product. Multi-level task scheduling processing technology solves the problem of the system cluster deployment and effectiveness.