<p>Characterization of the spatial distribution of geomaterials and of the associated attributes is a key step associated with the set up of a hydrogeological model. Geological information are often used as a basis for this purpose. One of the most common sources of geological information is provided by available borehole data. However, geological and hydraulic information are often available at different scale. In most cases hydraulic parameters are only measured at point locations, e.g. based on pumping test, which cannot be directly transferred into 3D large-scale parameter fields. However, in some regions even geological information are scare. In such situations, information about aquifer facies and material groups need to be interpolated and serve then as a base to derive key hydraulic parameters, such as hydraulic conductivity, or transport parameters, such as porosity, diversity or reactive surfaces. Sedimentary descriptions are usually achieved when drilling a borehole. Classification of sediments rests on a well defined procedure and provides a preliminary assessment on particle size distributions of the samples analyzed. Based on sedimentary descriptions of the boreholes we construct synthetic particle size distribution curves. These particle size distribution curves can be used to calculate major local attributes of the system (e.g., hydraulic and some specific transport parameters). Based on these types of readily-available information this study aims at developing a procedure to assist construction of a high resolution geological model suitable to be transferred into a flow and transport model that is then used for water resources management issues. We therefore aim to estimate storage and transmissivity with a high reliability by accounting for the material composition in the interpolated space. We rely on a compositional data analysis framework and represent particle size fractions associated with a given location as a compositional vector. These vectors are then projected onto a computational grid through compositional kriging to characterize the spatial heterogeneity of the system. We compare these results against an approach that is based on clustering the ensuing information to obtain distinct geomaterial classes and then assess their spatial distribution through indicator kriging. After the 3D field of grain size distribution curves is generated, they are transferred into hydraulic parameter. Although the process of clustering and using material classes is inevitably associated with a loss in information the procedure of forming a representative particle size distribution around the compositional clusters attempts to keep this loss of information at a minimum. The benefit of interpolating the compositional data instead of directly interpolating inferred parameters is that the particle size distribution curves contain a huge set of information from hydraulic to transport and reactive parameters, which would be lost using hydraulic conductivity exclusively, while the use of material classes increases the efficiency of the calibration of the groundwater model.</p>

<p>Groundwater recharge is an important variable for sustainable groundwater resources management in regions affected by water scarcity. The specifics of the Mediterranean require adapted techniques to also account for climate change implying a higher frequency of extreme events. Appropriate techniques are highly relevant for recharge with low rates. We compare three methods for the Western Mountain Aquifer, a karst in Israel: soil moisture budget calculations at basin scale, empirical functions, and machine learning algorithms. Resulting recharge are compared with measured spring discharge.</p> <p><strong>Neural networks</strong> have the advantage of not requiring much knowledge about physical processes or hydrogeological and hydrological conditions, nor about model parameters. This data-driven machine learning algorithms learn the non-linear relationship between precipitation events and spring water discharge given a sufficient amount of training data is available. After training, the neural network could be used as a nonlinear function to model recharge of any predicted precipitation time series. However, this approach does not allow for any quantitative analysis of external forcing, such as land use, or internal parameter, such as soil characteristics, nor does it account for any expected future change in precipitation pattern.</p> <p><strong>Hydro-pedotransfer functions (HPTF)</strong> are based on empirical relationships between precipitation and recharge. HPTFs account for potential evapotranspiration, annual precipitation, land cover, and a critical water supply (a threshold when actual evapotranspiration depends only on atmospheric conditions). Resulting percolation rates consider i) vegetation types, ii) precipitation during the vegetation growth period, iii) runoff, iv) plant available soil water, and v) capillary rise. The application of HPTF to a karst aquifer has the advantage that only limited input data are required. However, our results indicate that HPTFs are not able to capture the rapid recharge component observed in karst systems and thus underestimate recharge.</p> <p>The <strong>Soil Water Assessment Tool (SWAT) </strong>employs a hydrological and soil moisture budget calculations. Objective functions are actual evapotranspiration and surface runoff. Evapotranspiration is obtained from MODIS remote sensing data. Calibration of actual evapotranspiration is especially challenging for summer periods due to the impact of vegetation and irrigation. However, the most relevant parameter determining daily recharge rates are water loss by surface-runoff and surface water storage in wadi beds generating episodic recharge.</p> <p>Impact of shifts in climate is considered by climate projections obtained with the RCM COSMO-CLM at resolution of 3 km, under the IPCC RCP4.5 scenario, nested into the MENA-CORDEX domain. However, we believe that changes in land use from natural vegetation (trees, grass-, and shrublands) to rain-fed agricultural area could possibly shift the water budget from deficit to surplus conditions (recharge dominated). During the period 1992 to 2015 natural vegetation decreased by 8% and urban areas increased by up to 6%, while (rain-fed) agricultural areas remained almost constant. We investigate if land use changes might have (a much) larger impact on percolation rates than the predicted climate change effect. Thus, in future recharge may be controlled and enhanced in regions with water scarcity by better management of land use employing an optimized combination between precipitation, irrigation, and crop type.</p>

<p>Groundwater resources are expected to be affected by climate change and population growth and thus sophisticated water resources management strategies are of importance especially in arid and semi-arid regions. A better understanding of groundwater recharge and infiltration processes will allow us to consider not only water availability but also the sustainable yield of karst aquifers.</p><p>Because of the thin or frequently absent soil cover and thick vadose zones the assessment of groundwater recharge in fractured rock aquifers is highly complex. Furthermore, in (semi)-arid regions, precipitation is highly variable in space and time and frequently characterized by data scarcity. Therefore, classical methods are often not directly applicable.</p><p>This is especially the case for karstic aquifers, where i) the surface is characterized by depressions and dry valleys, ii) the vadose zone by complex infiltration processes, and iii) the saturated zone by high hydraulic conductivity and low storage capacity. Furthermore, epikarst systems display their own hydraulic dynamics affecting spatial and temporal distribution of infiltration rates. The superposition of all these hydraulic effects and characteristics of all compartments generates a complex groundwater recharge input signal.</p><p>Artificial neural networks (ANN) have the advantage, that they do not require knowledge about the underlying physical processes or the structure of the system, nor do they need prior hydrogeological information and therefore no model parameters, usually difficult to obtain. Groundwater recharge shows a high dependency on precipitation history and therefore the ANN to be chosen should be capable to reproduce some memory effects. This is considered by a standard multilayer perceptron (MLP) ANN, which uses a time frame as an input signal, as well as a recurrent ANN. For both large data sets are desirable. Because of the delay between input (precipitation, temperature, pumping) and output (spring discharge) signals, the data have to be analyzed in a geostatistical framework to determine the time lag between the input and the corresponding output as well as the input time frame for the MLP.</p><p>Two models are set up, one for the Lez catchment, located in the South of France, and one for the catchment of the Gallusquelle spring, located in South-West Germany. Both catchments aquifers are characterized by different degrees of karstification. While in the Lez catchment flow is dominated by conduit network, the Gallusquelle aquifer shows a lower degree of karstification with a stronger influence of the aquifer matrix. Additionally, the two climates differ, with the Lez catchment displaying a Mediterranean type of climate while the Gallusquelle catchment is characterized by oceanic to continental climatic conditions.</p><p>Our goal is to find neural network architecture(s) capable of reproducing the general system behaviour of the two karst aquifers possibly transferable to other karst systems. Therefore, the networks will be trained for the two different locations and compared to analyze similarities and differences.</p>

No abstract