There are currently many conflicts regarding water, including flooding, shortages, pollution, eutrification of static water, irrigation disputes, opposition to new dam constructions, ecological conservation of wetlands, and so on. To solve these problems, detailed information about the specific issues must be shared between the stakeholders, and Hydroinformatics is a major contributor to providing this information.
In Hydroinformatics, a hydrological model is the most important tool in enabling the sharing of water information. For this, lumped models (models that do not divide the total into sub-classes but use the same parameter values across the entire area) have been widely used around the world to describe the characteristics of river basins. However, the results of a lumped model, which are essentially averaged across the entire basin, do not fully satisfy the demands of the individual stakeholders.
To address this issue the distributed model is now being developed to furnish water information. In essence a distributed model consists of a subdividing of the area of study into sub-basins of tributaries and the main river channel. At the same time as the methodology of distributed modeling is developing, a number of commercial models are becoming available to support distributed modeling, to ease the modeling task for government institutes and non-governmental organizations (NGOs) and others. For this study, Wallingford Software’s InfoWorks RS was used.
The purpose of this study
Despite the progress of distributed river modeling, the methodologies are still developing, and the authors consider that a key topic to address is the size of sub-basins.
If sub-basins are large, the characteristics of the model are similar to that of a lumped model. With a distributed model, in principle the smaller the sub-basins the better for the level of improvement over the lumped model. But there are some problems with using small basins, related to the lack of availability of data at this level of detail – data on rainfall, runoff, and the cross sections of river beds.
Some existing research work addresses these issues. For example, the Coweeta Hydrologic Laboratory in North Carolina, US, is working with catchments of 0.09 to 7.60 square kilometers, and their results are useful in progressing distributed modeling.
However, there are still practical problems. When using small basins, the basic time interval of the simulation needs to be small, and specifically the time of concentration of the sub-basins. This has two impacts. The first is that rainfall data needs to be available with this time interval, which may not be the case. The second is that the amount of data and size of computation is large – a 5-day simulation of a model with 100 sub-basins with a hyetograph of 5-minute intervals means 144,000 units of rainfall data. This is a challenge for an ordinary PC. This is compounded by the fact that, given that in a large basin rainfall is not evenly distributed, so that some areas get heavy rain and some none, the best results require continuous simulation over several rainfall events and dry periods.
The purpose of this study is to examine the feasibility of continuous simulation in a distributed model with the smallest possible sub-basins, and to point out the requirements for further improvements.
Outline of the basin
The river basin modeled was the Kamitsue-mura and Nakatsue-mura in Oita Pref., Japan, an area of 182 square kilometers. The longest stream in the basin has a length of 20km, and almost all the area is clay loam and silty loam planted with cedar and cypress trees.
Hourly rainfall data was available from 12 rain gauges in and around the basin, and the hourly inflow to an artificial Reservoir was also used in the study. The PC used was a 3.4 GHz Pentium 4 running Windows XP, with 1GB of memory and 93 GB of disk. The software was Wallingford Software’s InfoWorks RS v6.0.
The outline of the methodology is as follows:
1. the main river channels were designated as objects of hydrologic routing.
2. the basin was subdivided into 129 sub-basins of average area 1.4 square meters.
3. runoff hydrographs were calculated from lumped models, the normal practice in a distributed model. This involved examining three components:
- surface runoff
- interflow (water that has penetrated the soil surface flowing between basins and into channels)
- base (or groundwater) flow
and the impact on these of:
- antecedent rainfall
- the increase in the runoff ratio during continuous rainfall
- the recovery of retention capacity of the soil in dry periods.
4. the SCS-CN surface runoff model was used, with amendments to compensate for the fact that it does not meet the needs of continuous simulation (because the S-value varies discontinuously with the antecedent moisture conditions in the model)
5. the interflow is an important aspect of Japanese mountain basins. The study team designed a tentative simple model for interflow, and adopted it for the sub-basins.
6. the unit discharge of the dry weather period in the basin was estimated as 1 cubic meter per 100 square kilometers from analysis of observed data, and that was added to the hydrographs of each sub-basin as the base flow
7. the key parameters needed for rainfall runoff models were estimated from the analysis of observed rainfall data and the flow data on the downstream end of the whole basin.
8. the hyetographs needed to be shortened from their base unit of one hour because this cannot describe the conditions of a small basin. The team converted the data to 5-minute intervals using quadratic and exponential curves.
9. the rainfall needed to be assigned spatially from the12 rain gauges to the 129 basins. Assumptions were made which effectively smoothed the rainfall data across the sub basins, using TIN (triangular irregular network) surfaces to obtain the smoothness and the vales for each sub-basin.
10. the river channels were estimated in 1 kilometer lengths with an assumed triangular cross section, and provided 100m transient sections where there were sudden changes.
11. with continuous simulation there are both wet and dry periods, and the latter can give rise to problems of low flow, and unsteady flow calculations, especially at convex points in the fall of the stream. To overcome this a 200-meter curve of stream bed was assumed rather than a convex point.
Simulation results and sensitivity analysis
The study team produced successful results for a continuous 5-day simulation, with generally good agreement between model results and observations. The basic concept of the study was confirmed by the results.
Sensitivity analysis showed that the potential maximum retention has a great effect on the peak flow rate, and that the value used in the study was effective. Leading to the conclusion that the model fairly reflects rainfall-runoff in this forested area.
Conclusions and further investigations
The study shows that the modeling software and an ordinary PC handled the data volumes of a large number of small basins easily. The next problem to address is providing data for sub-basins individually. This poses two requirements. The first is that the rainfall-runoff models must be prepared according to the characteristics of each specific basin. The second is that observations of rainfall and runoff must be conducted in small basins at 5-minute intervals to determine the parameters and their relationship to surface conditions.
Improvement of distributed modeling and in observation techniques are a double helix. The feedback between these two areas is the key issue in Hydroinformatics.
Tyrone Parkinson, Sales Manager for River Modeling at Wallingford Software comments:
"This is a very interesting study that highlights some of the key advantages of InfoWorks RS when applied to distributed models. The first key benefit is the fully integrated hydrological and hydraulic simulation capability of the engine. This allows the modeler the choice of using rainfall runoff, in-channel or overland flow routing or fully hydrodynamic representation of the differing parts of the watershed in a single network. This means that the modeler can concentrate the detail where needed, and where the data allows, while retaining the speed and data efficiency of traditional "lumped" approaches.
"Secondly, integration with the DTM enables rapid model construction by automatically extracting appropriate data for the selected modeling approach. The links with GIS also allow speedy assembly of rainfall runoff models from land use and soil type characteristics as well as selecting channel roughness values based on vegetation and bank cover.
"The fast and stable computational engine enabled lengthy simulations of both high and low flow periods to be run quickly, as the authors clearly note. And finally, the model management, version control and audit trail functionality aided the generation of the various scenarios for the sensitivity tests and provided a very user friendly and storage efficient environment within which to undertake the study and manage the results."