BTOPMC/SCAU分布式流域水文模型原理和系统设计

BTOPMC(block-wise use of TOPMODEL(TOPgraphic MODEL,基于地形的水文模型)with the Muskingum-Cunge method,分块应用TOPMODEL和马斯京根-康吉汇流方法)是一个具有较强物理基础的分布式流域水文模型,主要用于教学和流域研究,模型有待软件化,以便用于流域作业预报和水资源管理。该文采用多层架构体系设计了该模型的计算机系统,形成了BTOPMC/SCAU(south china agricultural university),系统由数据层、模型层、通信层、数据表达层和用户操作层组成,运行环境采用客户端/服务端模式。提供手动和自动2种参数率定模式,自动率定采用SCE-UA(shuffled complex evolution developed at university of arizona)全局最优算法,内置7种目标函数。采用Open MP(open multi-processing,开放式多处理)多核并发编程技术在产汇流模块实现了多核并行计算。系统为流域水文作业预报和水资源管理提供了精致的方法和简便工具。

Principles of BTOPMC/SCAU distributed watershed hydrological model with system design

Abstract: Based on a watershed digital elevation model (DEM) and by dividing a large basin into sub-basins or blocks, the TOPMODEL (TOPgraphic MODEL) concepts were applied for runoff generation and an adaptive Muskingum-Cunge method was proposed for runoff routing. A distributed watershed hydrological model was developed with a strong physical basis, named BTOPMC (a Block-wise use of TOPmodel with the Muskingum-Cunge method). Up to now, the model has been applied to watershed research globally and as a teaching tool of hydrological science in some organizations. In order to apply it to operational hydrological forecasts and water resources management over watersheds, a system named BTOPMC/SCAU (South China Agricultural University) was designed, consisting of 5 layers: data management, models, communication, data illustration and user operation. Data management layer laid at the system base, responsible for data storage and maintenance. Data was divided into two types: structured data and non-structured data respectively stored in relative database manage system and files. As a core of the system, models were consisted of modules of terrain analysis, runoff generation, flow concentration and basin application. The terrain module computed static characters of basin ground. The runoff generation and flow concentration modules computed dynamic hydrological processes by integrating meteorological inputs and basin ground characters. In order to improve modeling efficiency, an OpenMP (Open Multi-Processing) programming was used multiple cores of CPUs for parallel computation in the two modules. Based on an epoll mechanism and programmed in C/C++, the communication layer was designed for message passing among other layers and it supported simultaneous multi-users access. Depending on the user's intention it was passible after using some integration tools in the data illustration layer to extract inputs, outputs and some processing results from data layer and intuitively display them in tables or graphs. The user operation layer, which provided a concise GUI (Graphic User Interface), was programmed in Java, so it was able to run in different platforms such as Microsoft Windows, various Unix, Linux and so on. BTOPMC/SCAU was executed in a Client/Server environment where user operation layer and data illustration layer were deployed to clients and models and databases on server. The communication layer passed messages between Clients and Server. In this way, the system can concentrate on the huge burden of hydrological computation, and in the meantime allow large amount of data input and query from many users everywhere for the basin management. Two operation conditions were provided to run the system: calibration and simulation, and the calibration operation supported two methods: manual and automatic ways. In automatic calibration, a global optimization algorithm, SCE-UA (Shuffled Complex Evolution developed at University of Arizona), was used and seven kinds of objective functions ccould be chosen. BTOPMC/SCAU provided a set of sophisticated methods and convenient tools for hydrological forecasting and water resources management over watersheds. It was characterized with (1) A layered architecture but coupled weakly and allowed easy integration of advanced techniques. (2) Communicated in an epoll mechanism and supported multi-users accesses. (3) Modularized in model layer to include more methods and to develop more applications in the future. (4) Organization of data as non-structural (saved in files) and structural (managed in relative database) with high security, faster retrieval and storage, and accessibility. (5) Parallel computation dramatically enhancing calibration and simulation efficiencies in multi-core systems. (6) Multiple tools for visual queries, on-line statistical analysis, etc.