Problems with model selection arise not from problems with model availability, but with model suitability. By searching the literature and the libraries of different universities and research agencies one can find hundreds, if not literally, thousands, of models which have been designed to simulate hydrological processes and which help to better understand various problems related to water and land use management. In the end, the right selection will provide an invaluable tool for decision makers in finding out the optimum solutions to catchment management problems. To increase the usefulness of a hydrological model, it is important to choose the right model which will be able to fulfill the demands of the modelling application. This can be achieved by developing model selection criteria.
In Chapter Two, the different runoff mechanisms were discussed and the conclusions indicated that the infiltration excess overland flow is the dominant mechanism in the arid and semi arid regions. Nevertheless, saturation excess overland flow is also responsible for some runoff which is generated in the formerly mentioned areas. For simulating and quantifying these two very important processes, together with transmission losses and runoff routing, the physically based distributed models provides a good tool to achieve this task. So for Wadi Kafrein, the physically based distributed models showed an advantage compared with the conventional rainfall-runoff models, in order to understand hydrological processes and to analyze the rainfall-runoff generation mechanisms and amounts.
Although the traditional hydrological models of lumped conceptual type are well suited to deal with the main part of current water resources assessment and flood and drought forecasting, more advanced tools are required for the remaining problems including prediction capabilities. Distributed models are able to predict, therefore making them highly effective management tools (Refsgaard and Abbott, 1996).
The physically based distributed models present a better possibility in solving several hydrological problems, especially the ones from human impacts on land use change.
Such problems can be solved using the physical basis which takes into consideration the spatial and temporal distribution of physical parameters over the catchment. The data to be used in a physically based model needs to be prepared in a special way in which considerations of the physical properties of the catchment are included, meaning the spatial distribution of the data within the catchment need to be sufficient and representative. One of the strong qualities of using physically based distributed models is from the philosophy of the model use by estimating the necessary data and parameter values which are not available with only an approximate accuracy.
Furthermore, the physical basis of the model does not require a long hydrological and meteorological data series for calibration because the calibrated set of parameters should be applied outside the range of the conditions used in the calibration stage of the model (Abbott et al., 1986a).
The physically based distributed models can be, in principle, applied to solve all types of hydrological problems; they are based on our understanding of the physics of the hydrological processes and it uses within its source code physically based equations for processes description.
Based on the above discussion and by considering the objectives of this research, it has been recognized that a rainfall-runoff model with physically based distributed approach represents the best choice. Model selection tool criteria have been identified to choose the best model from the discussed models list in Section 3.4.
For the model code to be used in this study, the following minimum capabilities are required:
The ability of the model to simulate the hydrological processes: accurate partitioning of precipitation, evapotranspiration, runoff, and infiltration components is required. The simulation of the runoff component and related factors such as the runoff mechanism, transmission losses, and the routing processes are of special importance.
The ability of the model to deal with the prevailing climatic conditions and the hydrologic processes in arid and semi arid regions: represented by the high variations of precipitation and runoff distributions in space and time.
The physically based distributed structure of the model: flexibility in model application and capability for simulating hydrological processes efficiently and accurately.
The availability of the model for public use: availability in the public domain as an open source will be given a preference over the commercial models. The availability of documentation and support is desirable.
The availability of resources to meet the model input requirements: the required data should have the format, type, and time step, obtained by having first defined the available resources for this study.
The model should be able to fulfill the objectives of this study which have been mentioned in the first chapter.
A priority will be given for models which will be able to simulate the human effect on hydrological processes: the effect of urbanization on hydrological processes and the dynamic of water flow in the study area.
One cannot say that one model is better than another. Simply put, the development of the world‟s existing hydrological models has been borne out of the desire to fulfill many purposes: some models were developed with the aim of solving a particular hydrological problem, while others were developed for commercial purposes; while even some other models, which have been the topic of scientific papers, simply aim to simulate hypothetical scenarios in order to enhance our knowledge and understanding of our environment.
The models discussed in section 3.4 are only a few examples of models which have been developed and which are available. The empirical and lumped parameter models have been excluded in the consideration as a potential main model tool due to their structure which does not take into consideration the spatial distribution of the parameters. As discussed in Chapter 2, Wadi Kafrein has highly spatially variable soil properties, topography, slopes, vegetation, and precipitation, and therefore a model that does not take this into consideration simply does not meet the criteria for model selection. Therefore, the physically based models are the best available models for use in this study. The potential physically based models listed in section 3.4.3 will be discussed in more detail to select the suitable modelling tool to be applied in Wadi Kafrein.
DHSVM is a distributed hydrologic model that explicitly represents the effects of topography and vegetation on water fluxes through the landscape. The model has been applied predominantly to mountainous watersheds in the Pacific Northwest in the United States. During the literature review assessment of the DHSVM model, it has been found that most of the model applications can be arranged in three main categories as follow:
In hydrological forest managements; (i.e. Storck et al., 1998; Alila et al., 2001;
Thyer et al., 2003).
In land use and land cover changes; ( i.e. Burges et al., 1998; VanShaar et al., 2002).
In climatic changes impacts (i.e. Leung and Wigmosta, 1999; Wigmosta and Leung, 2001).
As stated earlier the model was applied predominantly to mountainous watersheds in the Pacific Northwest in the United States; these watersheds receive rain year-around and receives very heavy snowfall, which is known to be one of the snowiest places in the world. Therefore the hydrological responses of these watersheds are different compared with the arid and semi arid regions similar to Wadi Kafrein. Also no publications or reports were found which refers to model application or evaluation in arid or semi arid regions. Therefore the DHSVM was excluded as a main modelling tool for Wadi Kafrein watershed.
The MIKE SHE model and the developed model based on it; the SHETRAN, are both powerful models with several extensions and modules making them among the best available and long term developed physically based hydrological models. MIKE SHE covers the major processes in the hydrological cycle and includes process models for evapotranspiration, overland flow, unsaturated flow, groundwater flow and channel flow and their interactions. Each of these processes can be represented at different levels of spatial distribution and complexity, according to the goals of the modelling study, the availability of field data and the modeler‟s choices, (Butts et al., 2004).
MIKE SHE uses MIKE 11 to simulate channel flow. MIKE 11 includes comprehensive facilities for modelling complex channel networks, lakes and reservoirs, and river structures, such as gates, sluices, and weirs. Also the detailed model description and documentation with plenty of successful applications puts MIKE SHE on the top of potential models list. Nevertheless, the cost of the model is rather expensive.
For the aim of this study the TRAIN-ZIN model was chosen for the following reasons:
The availability of the model in the public domain
It is the only model among the potential models list which was developed and further enhanced to simulate the hydrological processes in arid and semi arid regions
The model simulates the runoff components in much detail with high spatial and temporal resolution
The model has several successful applications in arid and semi arid regions
The model can fulfill the objectives of this study and has the required capabilities mentioned earlier in this chapter
Further description of the TRAIN-ZIN model will be given in Chapter 5 which emphasize on the model application, parameterization, calibration and validation processes.
4 Methodology of data acquisition and analysis