Source–Pathway Characterization

The main difference between this approach and the previous one is that the transport mechanisms between sources and receptors or between inputs and outputs are explicitly modeled. This requires the relevant parameters to be quantified in order for the model to be calculated through for the entire process. There are many such process models in environmental science and engineering. In general, these either operate as lumped parameter or distributed parameter models, as discussed in Chapter 5. The lumped parameter models are either entirely nonspatial (e.g., the panda–bamboo interaction model in Chapter 5) or follow discrete spatial units, such as catchment areas.

A drainage basin, for example, can be partitioned into a number of subcatchments where the parameters are being lumped for each subcatchment. Thus, if one is considering the effect of several large drainage basins, one effectively ends up with a distributed lumped parameter model of some complexity. This is the case in the first source–pathway study to be presented where we will be looking at the coupling of GIS and hydraulic modeling for basin management planning. The second study will be a fully distributed parameter modeling of coastal oil spills.

Basin Management Planning

This case study comes from Hong Kong and involves the coupling of GIS (Genasys II) and hydraulic simulation modeling (MIKE 11) into a spatial decision support system (SDSS). Although such couplings are now commonplace, when this project was started in 1991, the approach was a novelty. Presented here are the aspects of the project and its background that are already in the public domain (United Nations, 1990; Brimicombe, 1992; Townsend and Bartlett, 1992; Brimicombe and Bartlett, 1993; 1996; Drainage Services Department, 2008). We have already seen in the previ-
ous chapter that Hong Kong is a small region of just 1050 km2 of which 60% is mountainous terrain and which, with a population of about 6 million, has resulted in intense development. Average annual rainfall is 2225 mm, but tropical depressions and typhoons can result in rainfall intensities that can reach 90 mm/hr. The steep terrain leads to rapid runoff concentration and flash flooding in lowland basins. In the period of 1980 to 1990, 84 major flood events occurred, many lasting two or more days. This compares with just 16 events in the 1960s, rising to 38 in the 1970s. Flooding occurs mostly in lowland basins and natural floodplains in the northern part of the region (Figure 6.5). This increase in the number of flood events reflects the changing land use of the region over this period. There are three key elements:

  • Urbanization: Building of new towns in low-lying and floodplain areas
  • Rural development: A combination of village expansion, the transformation of agricultural land to small industry and storage parks (containers, building materials, construction plant, and so on) and the abandonment of agriculture while speculating on development.
  • Floodplain reduction: The construction of ponds for fish and duck farming with bunds constructed to above flood level over substantial areas has lead to a reduction in floodplain area.

In 1989, in recognition of the growing problem, the Hong Kong government established the Drainage Services Department (HKDSD) to formulate a flood prevention strategy through basin management planning. These plans might include a mix of river training, flood storage schemes, flood proofing, and land use development control. The Town Planning Ordinance was amended to curb unauthorized development and required all new developments to undertake a specific drainage impact assessment. In 1991, a “Territorial Land Drainage and Flood Control Strategy Study” was commissioned and would form the basis for drawing up 1:5000 scale basin management plans (BMP) for each drainage basin. It was in this context that a coupling of GIS and hydraulic modeling into a SDSS first took place and was to become established practice in Hong Kong.

In this project, the decision to use hydraulic modeling dominated the architecture of the system. Hydraulic modeling is used to simulate runoff
along a drainage network in response to specified rainfall events. Models are constructed using true channel details (cross section, gradient), aggre-
gated runoff parameters, floodplain storage mechanisms, and significant structures, such as bridges, culverts, and weirs. Thus, hydraulic modeling
can be expected to give a truer simulation of channel capacities and their flows than hydrological modeling. Apart from studying the existing situa-
tion, hydraulic modeling is widely used as a design tool for remedial and mitigation measures and, therefore, can be used in “what if”-type analyses.
The simulation is based on nodes (modeling points) linked by successive reaches of drainage channels into a topological network. Nodes are usually
located at typical cross sections on a reach (rather than at a confluence) or used to represent locations of flood storage. Each node receives the cumulative flow from any upstream node and its own area of subcatchment.

Parameters describing the flow accumulation within the subcatchment are lumped and attributed to the node. Thus, although technically this could be described as a lumped parameter model, there may be 100 or more subcatchments in a typically sized drainage basin for Hong Kong. The choice of GIS software rested on two key factors. The first was the need to handle large data sets in a timely manner, which, given the power of PCs and their storage capacity in 1991, meant that realistically the software needed to run on a UNIX workstation. Second, as will become evident below, the software needed to be able to handle fully topological vector data, TIN structures, and raster data in an integrated way. At the time, GIS and hydraulic modeling were only loosely coupled in as much as they were run independently with only the exchange of data between the two using reformatted ASCII (text) files.

The overall pattern that developed was one of using GIS to integrate and preprocess data, which are passed on to the simulation and then to accept back the results of the simulation into GIS for postprocessing. By linking proprietary GIS and hydraulic modeling software in this way, the study team was able to assess flood hazard for current and projected land use scenarios over a range of rainfall events (1:2 through 1:200-year return periods) and for a variety of mitigation measures. On the basis of these multiple outcomes, well-founded decisions could be made regarding appropriate proposals and options for the BMP.