Search

Summary This paper compares two methods for modeling karst aquifers, which are heterogeneous, highly non-linear, and hierarchical systems. There is a clear need to model these systems given the crucial role they play in water supply in many countries. In recent years, the main components of soft computing (fuzzy logic (FL), and Artificial Neural Networks, (ANNs)) have come to prevail in the modeling of complex non-linear systems in different scientific and technologic disciplines. In this study, Artificial Neural Networks and Adaptive Neuro-Fuzzy Interface System (ANFIS) methods were used for the prediction of daily discharge of karstic aquifers and their capability was compared. The approach was applied to 7 years of daily data of La Rochefoucauld karst system in south-western France. In order to predict the karst daily discharges, single-input (rainfall, piezometric level) vs. multiple-input (rainfall and piezometric level) series were used. In addition to these inputs, all models used measured or simulated discharges from the previous days with a specified delay. The models were designed in a Matlab™ environment. An automatic procedure was used to select the best calibrated models. Daily discharge predictions were then performed using the calibrated models. Comparing predicted and observed hydrographs indicates that both models (ANN and ANFIS) provide close predictions of the karst daily discharges. The summary statistics of both series (observed and predicted daily discharges) are comparable. The performance of both models is improved when the number of inputs is increased from one to two. The root mean square error between the observed and predicted series reaches a minimum for two-input models. However, the ANFIS model demonstrates a better performance than the ANN model to predict peak flow. The ANFIS approach demonstrates a better generalization capability and slightly higher performance than the ANN, especially for peak discharges.

This work is aimed at estimating the transmissivity of highly fractured hard rock aquifers using a geostatistical approach. The studied aquifer is formed by the crystalline and metamorphic rocks of the Western Ivory Coast (West Africa), in the Man Danane area. The study area covers 7290 km 2 (90 km×81 km). The fracturing network is dense and well connected, without a marked fracture direction. A data base comprising 118 transmissivity ( T ) values and 154 specific capacity ( Q / s ) values was compiled. A significant empirical relationship between T and Q / s was found, which enabled the transmissivity data to be supplemented. The variographic analysis of the two variables showed that the variograms of T and Q / s (which are lognormal variables) are much more structured than those of log T and log Q / s (which are normal variables). This result is contrary to what was previously published and raises the question whether normality is necessary in geostatistical analysis. Several input and geostatistical estimations of the transmissivity were tested using the cross validation procedure: measured transmissivity data; supplemented transmissivity data; kriging; cokriging. The cross validation results showed that the best estimation is provided using the kriging procedure, the transmissivity field represented by the whole data sample (measured+estimated using specific capacity) and the structural model evaluated solely on the measured transmissivity. The geostatistical approach provided in fine a reliable estimation of the transmissivity of the Man Danane aquifer, which will be used as an input in forthcoming modelling.

A simple GIS-linked model for groundwater nitrate transport has been developed in the IDRISI GIS environment. The transport model, based on advection transport only, is directly incorporated into a GIS subroutine using the Pascal computing language. This model requires less data than classical approaches and provides a user-friendly model environment. The GIS-model was tested on a 20 km 2 hydrogeologic catchment, particularly vulnerable to agricultural nitrate pollution. The model was coupled with an unsaturated zone transport model (AgriFlux), which simulates water and nitrate fluxes leaving the root zone. The results indicated that the simulated nitrate concentrations were in good agreement with measured values. In order to compare the GIS-linked model with a more complete model, simulations were also performed with MT3D-MODFLOW. The similarities between the results of the two models confirm the validity of the GIS-linked model.

The purpose of the study is to demonstrate that correlation and spectral analyses can contribute to the regional study of a large karst aquifer. An example is presented for the La Rochefoucauld karst aquifer (Charente, France). Different types of spatially distributed time series provide valuable spatio-temporal information for the karst aquifer. The available time series consist of the spring flow rates, the flow rates at different locations in sinking streams, the piezometric levels, the electrical conductivity and temperature of the water, the atmospheric pressure and the precipitation. The analysis of the flow rates at the springs shows that the aquifer empties very slowly and has a large storage capacity. Hydrodynamic links were established between three of the four rivers flowing on the aquifer and the springs. The results also demonstrate the important spatial heterogeneity of the aquifer and indicate that the most rapid flow occurs in the northern part of the aquifer. Hourly piezometric and electrical conductivity time series indicate that the transmissivity of the aquifer varies when some conductive channels become desaturated during the low water period. The delays between the distributed recharge and the piezometric level, between the localized river input and the flow rates at the springs and between the electrical conductivities in rivers and the main spring provide information on the travel times in the aquifer. The observation of earth tides and barometric effects indicate that this apparently unconfined aquifer has a confined behaviour.

The Republic of Djibouti is the location of an exceptional geodynamic situation, the Afar Depression, which is an emerged triple junction of the Red Sea, the Gulf of Aden and the East African rifts, where a fault network is particularly well developed and controls the permeability of volcanic aquifers. Pumping test data are analysed for two major basaltic series, the stratiform series (3.4-1 Ma) and the Dalha series (9-3.4 Ma), which are distinguished mainly by their geodynamic characteristics and geological history although they are both recognized as trap rocks. Drawdown and recovery were studied by Jacob's logarithmic approximation of the Theis equation and by the Theis recovery method. Elsewhere, step drawdown data were interpreted with the Cooper-Jacob graphical solution and Rorabough's method. The results describe aquifers in the stratiform and Dalha basalts which are identified by different hydrodynamic characteristics. Stratiform and Dalha aquifer transmissivities range from 1.5 × 10 −4 to 5.7 × 10 −1 m 2 s −1 and from 2.9 × 10 −6 to 1.6 × 10 −2 m 2 s −1 . Storage coefficients determined from piezometer observations and computer simulations describe the aquifers in the Gulf basalts, a particular type of the stratiform series, as semiconfined, and those in the Dalha basalts as confined or unconfined depending on the area. Thus, the calculated specific capacity demonstrated that the stratiform basalt aquifers have more favourable hydrodynamic characteristics than the Dalha basalt aquifers. Specific capacities vary between 0.372 and 510.01 m 3 h −1 m −1 within the stratiform basalts and between 0.008 and 35.2 m 3 h −1 m −1 in the Dalha basalts. As they were originally composed of the same type of trap rocks, the stratiform and Dalha basalts should have a similar magnitude of permeability, and this was confirmed by some wells. The lower specific capacities exhibited by the Dalha basalts is explained by a longer period of alteration and greater hydrothermal activity, as has been observed in well cuttings and field studies.

The experimental site of Guilin is located near the village of Yaji at the border of the Lijiang Plain (Province of Guangxi, southern China). Within upper Devonian limestones and dolomitic limestones, a karst has developed in accordance with intense fracturing resulting from Mesozoic and Tertiary tectonic episodes. Its morphology is characterized by two major types of karst: a cone karst (peak cluster area) bounded by the Lijiang Plain which comprises a tower or mogote karst (peak forest area). Groundwater discharges from a perennial spring (discharge up to 7 m3 s−1) and three intermittent springs (discharge less than 5 m3 s−1). Some observation boreholes have been drilled on the peak cluster area dolines and on the Lijiang Plain in front of the spring. Tracing experiments in the dolines showed that the dolines are connected with the springs. The apparent flow velocity ranges between 4 and 180 m h−1. Analyses of storm responses of the perennial spring and the variability in time of 18O concentration, pointed out the important role played by the flows occurring in the subcutaneous zone of the peak cluster area.

A procedure to interpret hydrochemical data based on a combined use of factor analysis (correspondence analysis) and Piper's classification principle is outlined and applied to the study of a multi-reservoir hydrothermal system (Allevard) located in the French Western Alps. Values of physico-chemical parameters determined on 95 water samples related to a single point (Allevard Spring) and gathered from previous studies extending over several decades were submitted to this procedure. The approach proved successful in differentiating groundwater masses of different hydrogeological origin which mix and discharge at the outlet of the system. The statistical analysis, supported by structural and isotopic information, allowed formulation of a flow and mixing model in agreement with geological evidence and comprising near-surface, subsurface and deep subsystems.

The relationships between hydraulic (transmissivity, T ) and geoelectrical (transverse resistance, TR ) parameters in a strongly heterogeneous alluvial aquifer (Plain of Haouz, Morocco) are statistically analyzed. Samples of both parameters are large enough for this purpose. Several statistical regressions (linear, exponential, geometrical) of T on TR , together with several assumptions for the transverse resistance computation are tested. The best relationship is obtained when considering a geometrical regression of T on the TR of only the permeable beds. Additionally, when the values T i and TR i of a given pair i , are corrected to the same watertable elevation, the precision of the prediction of T is significantly increased. To conclude, a quantitative and reliable relationship between the two parameters is proposed to predict, at less cost and with an acceptable accuracy, the transmissivity of the Haouz groundwater.

The study of groundwater-table variations due to ocean-tide effects in a coastal aquifer may yield, under certain conditions, its hydrodynamical characteristics. Such a study has been attempted about a zone of the limestone aquifer of the Miocene of Algarve (Portugal). Two computation methods can be utilized, depending on the form of the free-water table evolution: (1) a Carson-Laplace transform, or (2) adjustment of a sinusoidal variation law to the fluctuations. Their application has permitted to characterize the aquifer's general hydrodynamical behaviour. The model that gives the most satisfying reconstitutions of the observations, particularly points out a seepage situation which fits quite well to the known geological conditions of the aquifer.