An integrated water-conveyance system based on Web GIS

Water resource shortages are a common problem in Beijing. Currently, water resources in Beijing are overdeveloped, as evidenced by shrinking wetlands, groundwater overexploitation and inadequate ecological water. Therefore, managing the city's water resources is very challenging. In addition to socioeconomic development and population increases, the shortage of water resources and deterioration of the water environment have become obstacles to sustainable socioeconomic development in Beijing (Gan et al. 2010). Thus, the south-to-north water-conveyance project is a foundational measure to solve the problem of water shortages. Additionally, other measures must be taken to broaden sources of income, reduce expenditures, handle sewage and better utilize flood resources (Li & Wang 2011). In 2008, the Beijing-Shijiazhuang emergency water supply project was completed and put into operation, and a total of 1.124 billion cubic metres of water flowed into Beijing from Hebei. Now, this project has become a new strategic water source in Beijing.

Hydrodynamic numerical simulation is an appropriate means for conveyance progress decision-making. Many unsteady flow simulation models, such as USM, CARIMA, CANAL, DUFLOW and MODIS can be utilized to compute the conveyance progress (ASCE 1993). For the Chinese south-to-north water-conveyance project, Fang et al. (2009) proposed an equivalent roughness method to simulate the effect of transportation structures in an open channel on the water level. Zhang et al. (2007) developed a one-dimensional (1D) numerical model in a large-scale conveyance channel with complex inner boundary conditions. Additionally, three commercially available software packages (i.e. CanalCAD, Mike11 and Sobek) were suitable for unsteady-flow simulation. However, all of these programs can acquire high-precision results for ordinary river-modelling applications but not for complex hydraulic conditions. CanalCAD is skilled in simple canals and theoretical studies and it is not able to handle branching networks, supercritical flow, and other situations that occur in practical systems. Though Mike 11 and Sobek improve the aspects of the above-mentioned cases in simulation, how to handle inverted siphon is not involved. Moreover, for the sake of modelling sluice, Mike 11 and Sobek have to calculate the canal with the sluice separately instead of integrating with the initial model. Furthermore, these software programs are incapable of establishing steady-state conditions (Clemmens et al. 2005).

As a large-scale inter-basin water-conveyance project, the Beijing south-to-north water-conveyance project has several unique characteristics. For example, it is a long-distance pipeline, with numerous types and amounts of hydraulic structures along its path. Sluices, outlets, inverted siphons, pumping stations, reservoirs and other hydraulic structures all play complex roles in controlling the flow. Thus, simulating long-distance unsteady flow is difficult.

Simultaneously, to share, understand, examine and compare various modelling conditions and outputs easily and interactively, a platform to access, organize, integrate, visualize and analyse these data must be established (Yang et al. 2013; Yu et al. 2014). Therefore, another goal is to develop a web-based platform to provide the system with spatial data management, analysis and display functionalities (Tsanis & Boyle 2001; Kumar et al. 2015). In recent years, the Web geographic information system (GIS) has been widely adopted for resolving the geographic related practical issue (Huang et al. 2016; Lee et al. 2016). Particularly, on the side of hydraulic and hydrological computer models, GISs are used for providing methods to publish and disseminate spatially related data via the internet (Choi et al. 2005). Several efforts have been made to demonstrate the applicability of utilizing GIS as an integrated environment for manipulating hydraulic and hydrologic data on the intranet and internet. Examples include Yuan's platform, which is integrated with Google maps and a hydrological model for the retrieval, analysis and visualization of spatially related data (Yuan & Cheng 2007); Jia's rainfall-runoff GIS system, which was developed based on the ARCGIS platform and a distributed hydrological model (Jia et al. 2009); and Yu's numerical model results post-processing system, which concentrates on the visualization of data computed by DHI software (Yu et al. 2012). However, most previous studies employed a mature GIS software package as the map server rather than independently developing such a server. Furthermore, only a limited number of studies have discussed this system's ability to integrate a hydrodynamic model with relatively large calculated quantities on the internet. In addition, although previous Web GIS platforms have made advances with respect to geospatial technologies and high-resolution data, because of model complexity, poor understanding of the underlying assumptions, tedious model calibration, and insufficient skills for handling geospatial datasets, such platforms remain difficult for stakeholders and local communities to fully utilize (Miller et al. 2004; Li et al. 2006; Ng Sandy et al. 2009; Kulkarni et al. 2014).

This article proposes the integration of a Web GIS server product (WGCS), which was developed by the authors following the Web Map Service (WMS), with a complex 1D hydrodynamic numerical model for the simulation of unsteady flows with complex inner boundary conditions for long-distance water conveyance. The model was developed to achieve continuous simulation of hydraulic transition processes in water levels. The discharge drastically changes according to the definition of the inner boundary conditions (e.g. sluice or inverted siphon). The integrated system is capable of forecasting and decision-making for the whole process of water conveyance.

The objective of WGCS is to provide a practical and easy-to-use tool based on WEBGIS for Beijing water conveyance decision support. The system integrates all the operation progress of the water conveyance simulation in the unified WEBGIS interface with minimum essential parameter inputs. Therefore, even a non-expert user can obtain reasonable and comprehensive results through the browser. To achieve this goal, WGCS including three parts (model setting, hydrodynamic calculation, result display) is implemented through independent development with proprietary intellectual property rights. In this way, the developed hydrodynamic model is more benefit to implement pre-process and post-process within the GIS interface due to the complexity of commercial packages in setting up the model, running simulations, and interpreting output. Though it is easy to gain results data from commercial packages, it is difficult to accomplish process data (e.g. model setting data and intermediate results). Furthermore, it is difficult to achieve intermediate results analysis during calculation, which is a significant function of the simulation system. Moreover, contrary to other large scale GIS systems, WGCS is devoted to a relatively light weight WEBGIS platform with less GIS function, so independent development did not give rise to too much increased workload. In contrast, it is simple to maintain and reuse the GIS system on account of independent development. Additionally, little modification was necessary to apply WGCS to other water conveyance projects. The model performance is demonstrated by applying it to routine work in Beijing. The WGCS can conduct large-scale hydrodynamic numerical calculations through web and visualized model data based on the vector map. To our knowledge, there are few similar systems that have been put into practice successfully.

The calculation engine of this system can simulate the water pipe network with a 1D hydrodynamic model. This model not only simulates the normal process of the unsteady flow but also the process in the presence of hydraulic structures (e.g. sluices, inverted siphons, pumping stations, and reservoirs). It is implemented using the OpenMP parallel language based on shared memory between multi-cores for accurate computation and fast convergence. In the study, based on the Saint-Venant equations, a model for one-dimensional unsteady flow coupling free surface and pressure flow is developed to simulate the water conveyance progress using a series of hypotheses. Free-surface-pressurized flow which exits in the long distance water conveyance progress in Beijing is difficult to be accurately solved by commercial software packages. Taking into account the structural characteristics of the pipeline, the staggered grid of discrete semi-implicit method is used for unsteady flow model. This can simulate the backwater effects and local head losses accurately (Li 2015; Li et al. 2015; Wang 2015; Wang et al. 2015).

To improve the stability of the model, a semi-implicit discrete method based on a staggered grid is adopted to discretize the model according to the characteristics of the drainage network. Pipe networks have several characteristics that distinguish them from river networks. First, the cross-sectional values (e.g. the radius of the circular pipe) and the slopes between every two nodes are constant throughout a pipe network. Second, the underground pipe network and ground are usually exchanged at a certain point. Thus, when the model is discretized, the discharge and cross-sectional figures are defined at the pipe section (element) centres, and the water levels or piezometric heads are defined at the nodes. The momentum equation is discretized at pipe sections (elements), and the continuity equation is discretized at nodes. This numerical discretization method makes data input more convenient and reduces the error of generalization and also increases the stability and ensures the conservation of the discrete format.

Sluices, culverts, under drains, tunnels, inverted siphons and other hydraulic structures are important with respect to the water-conveyance line. The key to describing the response characteristics of the channel is correctly simulating these hydraulic structures. However, water flow at the entrances of these hydraulic structures often exhibits strong 3D features and, thus, does not satisfy the aforementioned Saint-Venant equations. Therefore, it must be generalized, treated as a boundary condition, and then solved simultaneously with the discretized Saint-Venant equations.

Sluices are used to regulate the flow over the water-conveyance line. The water level can be controlled and water discharged by varying the size of the sluice opening. Because several flow regimes may arise, such as free orifice flow, submerged orifice flow, weir free flow, and weir submerged flow, the flow model must be able to describe not only the various flow regimes but also the flow transition process caused by water flow changes. Clearly, this requirement increases the difficulty of flow simulation.

During the simulation, the corresponding head loss between the start and end should be calculated according to the discharge of the inverted siphon. Then, by solving discrete Saint-Venant equations, the water levels at the entrance and exit of the inverted siphon can be obtained.

In this paper, a new Web GIS application solution, which adopts C#, ASP.NET, Silverlight and database technology to overcome the drawbacks of traditional GIS applications, is designed. The traditional desktop GIS system is not widely used because of drawbacks, such as the requirement for expensive GIS software, the dependence on a specific operating system, the isolation of different computers, unfriendly user interfaces and difficulty with interactivity. Thus, Web GIS (a GIS based on the web) can overcome these limitations. Thus, the user can access the application or implement model simulation through a browser, and the equipment is inexpensive, regardless of the client's operating platform, and can be used without installing any GIS software. The benefits and advantages for data accessibility and analysis are obvious. Data are stored in a database, and thus, anyone who is authorized can access the shared data. In addition, the well-designed interface offers the user an interactive, flexible, and fast communication experience.

The standards which are most widely adopted for web based GIS tools have been released by Open Geospatial Consortium (OGC). Three implementation criterions of OGC standard (Web Map Service (WMS), Web Feature Service (WFS) and Web Coverage Service (WCS)), which grows in popularity and application, afford a basic infrastructure for sharing and switching the data across the internet. WMS, WFS, WCS returns Map, KML and data respectively according to request.

The new Web GIS server platform is accomplished based on WMS. Two necessary requests (GetCapabilities and GetMap) of WMS are implemented for publishing the geo-referenced map on the web. The former one is employed to return an XML file describing metadata. The latter one is used to return available layers, spatial scope, and format of image. The implementation of WMS only facilitates the image web mapping or cartography excluding model data plot.

Hydrodynamic model data (e.g. layout of pipeline or mesh) exhibit similar geo-spatial characteristics as the GIS data. Model results data which are attached to the mesh also demonstrate geographic attributes. Despite all this, considering that spatial data and model data are discriminative with their accessibility and confidentiality, model plots layers (vector layers and animation layers) are appended to the map as an affiliated layer based on Silverlight.

Image registration is usually the first fundamental problem that must be solved for the completion of raster-layer rendering. In this study, rectangular images corresponding to mappable units are downloaded from Google and employed for map rendering. The batch registration method is utilized to overcome the shortcomings of single registration using other image-registration tools. The registration process is performed in two steps. (1) Define the coordinate system. The geographic coordinate system should be transformed into a projected coordinate system because different coordinate systems are used in Google images and the numerical model. Additionally, because of the different coordinate system definitions used by Google and the Silverlight painting system (in which the origin of the coordinates is the upper-left corner, the X axis is positive to the right of the origin, and the Y axis is positive below the origin), another coordinate transformation must be performed: from the Silverlight coordinate system to that used by Google Earth. The new coordinate system is considered the global coordinate system for the GIS platform in this study. (2) Upload the raster data. Raster data are returned to the client in the form of a byte array, which is subsequently converted to the BitmapImage class. Then, the raster is translated to the specified location and added to the raster layer.

Rendering a static vector layer is simple because Silverlight offers a number of convenient and practical vector-drawing functions; the EllipseGeometry and LineGeometry classes are applied for drawing points and lines, respectively, in the global coordinate system defined above. Consequently, several attributes are defined, such as Stroke (describes how the shape outline is painted), StrokeThickness (describes the thickness of the shape outline) and Fill (describes how the interior of the shape is painted). Finally, the path class is adopted for adding a vector to the specified location of the specified layer. The efficiency of the painting system is relatively high because all the vector units of one layer are rendered as a single object.

The animation layer is rendered by alternately rendering a static flow field over a certain time interval according to the time sequence. The process of drawing the animation layer is as follows:

• (1)

  Initialize trace particles. To describe the particle movements and particle positions after ⁠, the concept of identity distance is introduced, which is defined as the distance between the first point of the element where the particle is located and the particle position. At the initial time, trace particles are arranged at the centres of the elements; thus, the identity distance of each particle is equal to half of the length of the corresponding element, and the physical values (e.g. water level, discharge and velocity) at the trace particles are equal to the physical values at the elements.

• (2)

  Compute the positions of the particles. The movements of the particles can be divided into three types: to the next element, to the front element, and in the current element.

  Initially, a particle's moving distance can be calculated as:

  

  (6)

  where is the flow velocity, and is the time interval for animation.

  

  where is the identity distance at time t. is the distance of the element where the particle is located. When the particle flows into the next element, and the adjacent elements of the current element must be determined. If it is adjacent to a single element, the identity distance will be calculated as ⁠. If the current element is adjacent to multiple elements, delete this particle and supplement particles at adjacent elements where no particle is located. When ⁠, the particle flows into the front element, and the adjacent elements of the current element must be determined in the same way. If it is adjacent to a single element, the identity distance will be calculated as ⁠. If the particle remains in the current element, the identity distance will be calculated as ⁠. If the current element is adjacent to multiple elements, delete this particle and supplement particles at adjacent elements where no particle is located.

  If the time step is so large that the particle is already flowing out of the next element or the front element, the particle must arrive at the current element via linked elements. Therefore, the particle is thought to rest on the next element's node; as a result, visually, the particle always flows along the river.

  The particle's coordinates are easily computed by linear interpolation as follows:

  

  (7)

  where is the particle's identity distance, is the element's location ⁠, and is the location of the element in which the particle is located.
• (3)
  Compute the particle's physical values. Time interpolation, rather than distance interpolation, was adopted for the calculation of the particle's physical values and is computed as follows:

  

  (8)

  where t is the current time during animation; and are previous and next times, respectively; and and are physical values at the previous and next times, respectively.
• (4)

  Paint the particle. The method used to paint the particle is similar to the method used for node painting. Note that the radii of the particles can be used to represent the magnitudes of the variables.

• (5)

  Animation. In the Silverlight software, the DispatchTimer class is employed to replace the traditional Timer class as the background thread timer. The DispatchTimer is the authentic background thread, which is performed independently rather than in the UI thread; therefore, DispatchTimer offers a significant advantage.

The usage of the DispatchTimer class is relatively simple. First, set the Tick event and interval attribute. Then, start the DispatchTimer, and perform the drawing program as described above. Finally, stop the DispatchTimer when the animation is complete or interrupted by the user.

The two modules discussed above (i.e. the hydrodynamic model for water-conveyance simulation and the Web GIS platform which is responsible for modelling and data analysis) are the main components of the entire system.

Because the procedure of model execution is distinct with no GIS capabilities requirements, the adoption of a loosely coupled method for integrating the two modules can sufficiently achieve the objective that all the procedures are fulfilled within the GIS interface. The hydrodynamic model module, which was developed in the FORTRAN language, is encapsulated in a separate executable program that waits to be called. The standard output stream from FORTRAN is read by C# to acquire the progress of the computation as it occurs. Moreover, visualization of the computational progress is implemented by reading partially output results saved by FORTRAN in a text file. After the simulation ends, data from the text file will be translated to the database for further analysis.

This loosely coupled method reduces complexity and dependency of the overall system. Interaction between modules occurs through data only, program update or extensions inside each module is not affected. So it is convenient to debug and test program. In addition, it is liable to apply the modules to other applications.

Note that only a single user is served through the first-come, first-served mechanism when simulation service is requested. The adoption of a parallel algorithm for a single user ensures that the calculation speed is relatively fast.

The design and development of the system follows the rules of modular software design and the reuse of open-source projects, so that the system obtains the features of economical extendibility, open interfaces, easy accessibility and low maintenance. Figure 1 illustrates the architecture of the system from a logical perspective.

The data platform is used to store the data that support the entire system, including the attribute, spatial, graphical and model databases. In addition, and most importantly, the database must be able to store large computational datasets for different simulation scenarios.

The service platform consists of the professional hydrodynamic models, Windows Communication Foundation (WCF) Rich Internet Applications (RIA) services, and GIS Services. WCF RIA services are primarily used to copy data from the SQL Server database to the web for data query, retrieval and analysis, while also providing data to the GIS Services used for spatial and computational data access and processing. Users can customize the prediction schemes for various scenarios on the service platform.

The Web platform mainly consists of the Silverlight App embedded in an Asp.Net web page. It mainly provides data query, browsing, analysis, statistics and other functions for the attribute, spatial and computational result data. It is also used to develop GIS Services, map computing nodes, and map discrete elements, among other features.

The user platform is also called the human-computer interaction layer. It is expressed as a Web browser, to provide users with a website operational interface.

Despite many obvious advantages of Geodatabase (ESRI) and oracle spatial (Oracle) in GIS development, but the cost is very high. Though PostgreSQL with the PostGIS is open source DBMS, which can reduce the cost, the adoption of PostGIS must make extra work for data conversion since there are already numerous spatial data (e.g. topography near the pipeline, distribution of river and lake) stored in the SQL Server database. Considering multiple factors such as cost of commercial software, cost of maintenance and requirements of proprietor, SQL Server 2008 is selected instead of Geodatabase as the database platform which is divided into four parts (Figure 1). Attribute datasets defines the feature of the modelling area. Spatial datasets stores geo-referenced information (e.g. spaces coordinate reference, extent of modelling area and geometry constitution). Graphics datasets records images of modelling areas as the background map. Model datasets defines descriptive information for numerical model configurations and simulation results.

Because of space constraints, we take the model database design as an example to illustrate the structure of the database. In this study, a model database is defined as comprising of descriptive information for model configurations and simulations.

Figure 2 shows the overall database structures of the model database, which includes the initial, computational and result data. The table RiverDiversion is the main entry point for initial data, at which a unique ID is auto-generated for indexing and identification. It contains brief information concerning the study area (e.g. the river name and river length). The detailed parameters, such as the coordinates of the identity points for the river description, are stored in the table RiverDescription. The ‘one-to-many’ relationships are established between RiverDiversion and RiverDescription. The table CrossSectionPosition stores the measured cross-section in terms of its relationship to CrossSectionFigure, where the parameters (e.g. initial point and elevation) for cross-section definition are stored.

The tables used for computation mainly store model data and computational parameters. The tables RiverNode, RiverElement and RiverCrossSection store discrete nodes, discrete elements and element values, respectively, in terms of their relationships to parameter tables for other conditions (e.g. NodeInitialH, where the initial node number and water head are stored; and BoundaryCondition, where the node number, boundary condition type, boundary condition unit, and boundary condition progress are stored). Hydraulic structure parameters are stored in the tables Reservoir, WaterCompany, PumpStation, and Sluices. Each table contains information regarding the element numbers or node numbers where the hydraulic structures are located and the characteristic parameters of the various types of hydraulic structures.

The result data tables are HydraulicElements and SluicesOpening. The table HydraulicElements has a ‘many-to-one’ relationship to RiverElement, which contains the physical variables (e.g. water level, discharge and velocity) simulated in the numerical model. It is filled in automatically when the simulation is complete. Another important piece of result data is the opening of the sluice, which is stored in the table SluicesOpening.

This system can be divided mainly into the GIS interface, the hydrodynamic model and database management. The integration between modules is achieved by adopting a coupling method so that model operations are fully covered within a unified GIS interface. Figure 3 presents a conceptual diagram of the GIS-model integrated system. The GIS interface is responsible for map publishing by layer, as described above; data visualization; and simulation implementation. Because of the adoption of role-based access control in this system, advanced users have permission to set up the model or update the model data, whereas ordinary users only have permission to simulate the scheme by inputting different condition data. Data management is conducted using a database that furnishes computational data to the hydrodynamic model and receives result data for GIS demonstration.

The procedural implementation of the integrated system consists of three sequential steps: (1) pre-processing, (2) model execution, and (3) post-processing procedures. Original data are read in a specific format, and the DXF file is converted into computational data by the model setting module, which includes a distance function control algorithm for meshing and a series of interpolation algorithms. Model and condition data are stored in the database using standard SQL statements or stored procedures (deployed in DBMS). Because of the inadequacy of FORTRAN for database operations, a data exchange interface between text files and the database had to be established to integrate pre-processing and post-processing data. The detailed data-processing procedure is illustrated in Figure 4.

The entire water-conveyance system, which combines pressurized and open channels, is divided into three parts: main-line projects, storage projects and water companies. The main-line projects are 187 km long and include the south main line, the east main line, and the lines from Tuancheng Lake to the ninth water-company conveyance project and the Miyun reservoir conveyance project. There are five storage projects: Daning reservoir, Daning surge tank, Miyun reservoir, Tuancheng Lake surge tank and Yizhuang surge tank. Currently, six water companies are in service, in addition, 15 water companies are under construction or undergoing expansion.

The main-line project includes 10 single projects, of which Huinanzhuang -Daning surge-tank conveyance pipeline (PCCP) has a diameter of approximately 4 m, with pipes running side by side.

A comprehensive demonstration of the Beijing conveyance pipeline by the Web GIS system is presented in Figure 5 and includes a map window, a model-setting window, and an operation and display window. The model-setting window lists the steps for model setting. The operation and display window, which comprises three tabs, indicates the simulation progress, relevant operations, and input parameters. The map window shows a zoomed map of the Beijing conveyance pipeline, overlaid with several vector layers (e.g. nodes and elements) representing the numerical computational elements. The flow animation layer in the map window, defined by running trace particles along the pipeline, illustrates the magnitudes of the variables and the flow velocity directions and magnitudes. The identity results dialog box shows the information in the form of time series at element source number 1855, which is highlighted in red in the GIS map.

It is worth noting that the webpage is using localhost on port 2844 as the server on the same machine as the browser. The practical water conveyance system is a Chinese version and the English version is in the debugging stage. Figure 5 is a screenshot of the development.

The administrator takes charge of modelling and parameter calibration which is implemented through a Web GIS window. The modelling procedures contain the following three steps. (1) Acquire original data from DXF file, which contains two layers, one is description of the pipeline in form of ‘polyline’, and the other is the measured cross-section in form of ‘point’. Next, acquire a detailed cross-section parameter corresponding with the point of the measured cross-section. (2) Generate a relative uniform mesh based on interval functions according to ‘Interval Length’. Consequently, several interpolation methods are employed to ensure that every divided element characteristic is computed. (3) Input initial water lever, initial discharge and roughness. Parameters defined at each element and node can be set respectively.

The model is generalized to 128 pipelines, which are divided into 2052 mesh nodes and 2080 elements. The length of each element is approximately 200 m. To describe the characteristics of the pipeline, 165 measured cross-sections (i.e. pipes, culverts and open channels) were defined. Using the interpolation algorithm in the model-setting module, the cross-sectional figures of each element can be calculated. The computed parameter values are listed as follows: The initial water level is 49 m, the discharge is 0 m²/s, and the roughness is 0.011–0.017. The progress of the water level or discharge is defined as a boundary condition at the entrance, and the relationship between the water level and discharge is defined at the water company. Furthermore, the hydraulic structure parameters are defined as shown in Tables 1–3.

When model establishment is completed, the ordinary user is responsible for simulation of water conveyance. Currently, the Huinanzhuang pump station is not pressurized, and thus, the current status maintains a small artesian flow. This case study simulates the operational conditions on September 12, 2015, at 8:00 using the hydrodynamic model described above. The operation of the scheduling process is illustrated in Figure 6. After a summary of parameters input and evaluation for water supply and water requirement, simulation is running by ‘Remote service’ button. The simulation status is indicated through a message bar and the two important outputs (water statistic and scheme order) of the WCGS tool are generated at the end of the simulation. Detailed analysis can be carried out on the map (e.g. time series or animation).

The elaborate analysis of all the results data is shown in Figure 6 and Table 4. Figure 7 demonstrates the overall water conveyance process. The water level at the Huinanzhuang forebay is 60.69 m, and the discharge from the main line is 13.32 m³/s. The water level at the PCCP pipeline starting point is 58.601 m. The inflow assigned to the third water company is 0.72 m³/s through the main line; that assigned to the urban river and lake is 8.29 m³/s through the Tuanchenghu Lake recession entrance; that assigned to Guogongzhuang water company through the south main line is 2.43 m³/s; and that assigned to Daning reservoir has a resting discharge of 1.88 m³/s through the Daning canal. The initial water level at Daning reservoir is 50.48 m, and the initial water level at Tuancheng Lake is 50.48 m.

Table 4 illustrates the comparison between modelled discharge or water level and observed discharge or water level for the relative engineering. The observed data are collected from the report that summarizes the actual measurement data of hydraulic elements by Beijing in China. The comparison was made for a total of five engineering cases. As summarized in Table 4, there was no significant difference between model results and the observed data. Therefore, the hydrodynamic model is clearly capable of simulating long-distance water-conveyance projects involving many types of complicated hydraulic structures.

To demonstrate the ability of the conveyance model integrated with the Web GIS platform to guide daily conveyance work, data observed from 08:00 on January 1 to 12:00 on August 14 (5,412 hours) were employed for model verification. The scheduling process involved three different stages during this period: small artesian flow (13 m³/s), large artesian flow (24 m³/s) and pressurized flow (40 m³/s).

The total inflow from Huinanzhuang is assigned to the third water company, Tiancunshan, and Chengzi through the main line; to the Guogongzhuang water company through the south main line; and to the Daning reservoir through the Daning canal or the urban river and lake through the Tuanchenghu Lake recession entrance. The scheduling processes for different sluices are fairly similar, as is normally the case.

According to the hydrodynamic numerical model, by using the observed water demand of the water company and the water level of Huinanzhuang forebay as boundary conditions, the water-conveyance process was simulated from 08:00 on January 1 to 12:00 on August 14 (5,412 hours). From 0–1,788, 1,788–4,563 and 4,563–5,412 h are small artesian flow (13 m³/s), large artesian flow (24 m³/s) and pressurized flow (40 m³/s) respectively. Whether the water levels of the control points (i.e. the Daning reservoir, Daning surge tank and Tuancheng Lake) meet the requirements of the control water level was also determined.

Model verification is described in Figures 8 and 9, which show the water level and discharge distribution respectively. Water levels at Daning surge tank and Daning reservoir were employed for comparison. Daning surge tank and Daning reservoir are significant regulating-and-storing engineering for water transferring security. The Xisihuan pipeline and loop line were adopted for discharge comparison. Generally, the modelled water level and discharge were in good agreement with observed data, except that particular points were underestimated or overestimated. This may be caused by inappropriate parameters for hydraulic structures or uncertain measurement errors. However, the agreement between modelled data and observed data is acceptable.

To show the magnitude of the difference between calculation results and observed data, absolute percentage error to water lever or discharge computed by (|measured-modelled|/measured) × 100% is summarized in Figures 10 and 11. From Figure 10, absolute percentage error to water level was generally small and most of them in Daning surge tank and Daning reservoir were lower than 2%. However, Figure 11 shows that a larger absolute percentage error, which was heavily concentrated in small artesian flow, was generated. Absolute percentage error in the South main and lift line of Xisihuan was lower than 10 and 19% respectively. In the right line of Xisihuan, there was a series of noticeably large values being higher than 50%, but the mean absolute percentage error for the right line of Xisihuan was about 16%. This may be caused by the difference between modelled parameters and actual parameters (e.g. flux coefficient) of No. 2 sluice.

Hence it is certified that the hydrodynamic model was reliable and practical for normal water-conveyance process. Furthermore, the model verification case demonstrates that the model of the complex hydraulic structure was adequate to simulate the pressurized flow and free surface flow over long distances and long durations.

Due to the complexity and synthesis of water conveyance, in order to improve the accurate of the simulation, the following work may be carried out in the future. (1) To enhance the practicability of the hydrodynamic model, other hydraulic structures such as drain well, exhaust valve and weir should be hypothesized and summarized. (2) Uncertainty analysis should be discussed to evaluate how the uncertainties will affect model results.

Because hydraulic transient process simulation is an essential issue, integrating hydrodynamic numerical models into a user-friendly GIS platform for simulating conveyance scenarios should be beneficial for routine work. As presented in this paper, the Web GIS-based conveyance system (WCGS) comprises pre-processing, hydrodynamic simulation and post-processing steps integrated via a Web GIS platform, which provides easy access to information and interaction. Once the system administrator sets up the hydrodynamic model for the conveyance pipeline, even a non-expert user can run simulations for different conveyance scenarios through a web browser using an ordinary computer. Furthermore, the application of the developed model was demonstrated for two cases of scheduling processes in Beijing, China. The case of the small artesian flow revealed that the conveyance order generated by the WGCS meets the requirements of the water diversion scenarios and control water level. The verification case simulated small artesian flow, large artesian flow and pressurized flow. The agreement between the numerical results and measured data confirms that decisions made based on the WGCS are scientific and reasonable. The results of the case studies illustrate that the WGCS can be used as an effective and practical water-conveyance simulation tool.

This research work was supported by the National Key Research and Development Program of China (Grant No. 2016YFC0402707), the National Natural Science Foundation of China (Grant No. 51309052), the Fundamental Research Funds for the Central Universities (Grant No. DUT15LK01), and the Fundamental Research Project of Key Laboratory of Liaoning Provincial Education Department (Grant No. LZ2015012).