Investigation of parameterization strategies for the hydrogeological module of Hydrogeios software - Application to the Boeoticos Kephisos basin

Boeoticos Kephisos river basin, located at the Eastern Sterea Hellas region, is the study area of the current thesis. The Boeoticos Kephisos river stems from the Mt Parnassos and flows into the Lake Hylike, forming a hydrographic network which extends to the Fthiotida, Fokida and Boeotia counties. The geological basin of the area consists of heavily karstified limestone, mostly developed on the mountains, flysch and alluvial deposits, lying in the plain areas. Due to its hydrogeological features the area is characterized by a considerable groundwater yield. The main discharge points are springs, located at the middle and upper parts of the basin. A considerable amount of water is also discharged into the sea, which is difficult to quantify. Several studies have been performed in the Boeoticos Kephisos basin, regarding its geological and hydrogeological features, while several models have been proposed in order to simulate the hydrological cycle, e.g. the study performed on behalf of the Hellenic Ministry for the Environment, Physical Planning and Public Works, research work based on MODFLOW which was performed by Nalbandis et al. (2000, 2002) and the model developed by Efstratiadis et al. (2006) and Efstratiadis (2008) based on HYDROGEIOS. The current thesis aims to extend previous studies on the area. The prime objective of this study is the development of a model, which will represent all the hydrosystem procedures of the Boeoticos Kephisos watershed. The model development is based on two simple principles: ensure consistency with the physical characteristics of the system, while keeping the number of parameters as low as possible. The GIS-based computer software HYDROGEIOS is used, which provides a holistic framework combining both hydrological and hydrogeological data of a river basin. The HYDROGEIOS modeling framework has the ability to incorporate decision-dependent abstractions and interactions between surface and groundwater flows, on the basis of a semi-distributed configuration. A semi-distributed approach and a monthly simulation time step are adopted. Multiple levels of schematization and parameterization of the surface system are utilized, by combining multiple levels of geographical data. The study area surface is divided in thirteen sub-basins (compared to five in Efstratiadis, 2008) and as a result more branches from the river network are considered. The groundwater flow field is divided in 40 irregular shaped cells. Four of these cells are located outside of the river basin to simulate underground leakages to adjacent basins or to the sea. Several factors are considered in the cell formulation. An effort is made to maintain homogeneity within the cell, regarding its hydrogeological features and avoid abrupt slope changes. The spatial distribution of the springs is also taken into account, as they usually indicate a change in the physical characteristics and groundwater flow. Using a fine cell system would lead to a large number of parameters. Consequently, the intergraded surface water – groundwater approach applied in the current study, would lead to time consuming and computationally intensive simulations. To avoid this problem, a permeability based cell grouping is adopted. Three groups are finally formed (permeable, impermeable and semi permeable) and a unique value for the hydraulic conductivity and the porosity is implemented in each one of them. For the aforementioned schematization, a total number of 64 parameters is required (16 for the groundwater system). A model calibration is then performed, based on monthly precipitation and potential evapotranspiration data of a five-year period. The models are validated on the following five-year period. Computed data at six control points (spring locations) and at the basin outlet are used to calculate several performance criteria. The hydrographs from the computed data are compared to the corresponding hydrographs from the measured data. Due to the complexity of the model and the multiple sources of uncertainty, it was not possible to achieve credible results in a few runs and the final result was reached gradually: starting from a relatively good solution, the number of the unknown parameters is reduced, in order to focus on those of significant importance or uncertainty. User intervention proved to be of critical importance during the calibration procedure, since small changes in the boundary conditions of the cells or the water levels improved model performance and its predictive capacity. In each run an optimization module, based on the evolutionary annealing-simplex method is implemented. Regarding the simulation results, the model developed in the current study provided satisfactory performance in most regions of the Boeoticos Kephisos basin. Increasing the model performance, while ensuring model parsimony and consistency with physical processes was a prime objective of the current thesis, but it was proved a rather difficult task. Specifically, the difficulties in quantifying the parameters and boundary conditions in the cells which contain springs amplified the model uncertainty and highlighted the high complexity of the karst system. Overall, due to the size of the study area and the number of the related parameters the model is vulnerable to various uncertainty sources and as a result the problem becomes ill-posed. To reduce the degree of uncertainty, an effort was made to incorporate all available data in the model through the appropriate parameters. Further analysis on daily basis would provide more accurate results, while sensitivity analysis of the parameters would assist the researchers to focus on factors of significant importance.