Evaluation of fish habitat suitability using a coupled ecohydraulic model: Habitat model selection and prediction

The selection of an approach to evaluate habitat suitability for a specific fish or life stage has been a matter of concern in habitat quality modelling studies. This study has taken Jinshaia sinensis, a commercially valuable fish endemic to the Jinsha River, China, as the target fish species. One‐ and two‐dimensional hydrodynamic models were coupled and combined with fish habitat models for a middle reach of the Jinsha River. The resulting ecohydraulic model was used to predict the changes in hydrodynamics and spawning habitat suitability that resulted from the operation of an under‐construction reservoir downstream of the study area. The preference function (product, arithmetic mean, geometric mean, and minimum value) and fuzzy logic habitat evaluation methods were compared to predict the spawning habitat suitability of the fish. The model was validated using the numbers of spawning eggs, and the results show that both the arithmetic mean and fuzzy logic method can be used to predict spawning habitat suitability. The model predictions show that the hydrodynamics of the study area would be altered if the impoundment water level exceeded 969 m. During the spawning season, the spawning habitat suitability would increase from April to early June and has little change from early June to July under the impact of the reservoir impoundment. The optimal river discharge rate for fish spawning is ~3,500 m³/s, and this would not change after the reservoir begins operation. This research can benefit other regions that will be affected by planned dams by predicting the impacts of reservoir operation on fish habitat quality, and the results will help decision makers protect the health of rivers and the overall ecosystem.     

Water Resources Assessment at Piracicaba,Capivari and Jundiaí River Basins: A Dynamic Systems Approach

The Piracicaba, Capivari, and Jundiaí River Basins (RB-PCJ) are mainly located in the State of São Paulo, Brazil. Using a dynamics systems simulation model (WRM-PCJ) to assess water resources sustainability, five 50-year simulations were run. WRM-PCJ was developed as a tool to aid decision and policy makers on the RB-PCJ Watershed Committee. The model has 254 variables. The model was calibrated and validated using available information from the 80s. Falkenmark Water Stress Index went from 1,403 m³ person^???1 year^???1 in 2004 to 734 m³ P^???1 year^???1 in 2054, and Xu Sustainability Index from 0.44 to 0.20. In 2004, the Keller River Basin Development Phase was Conservation, and by 2054 was Augmentation. The three criteria used to evaluate water resources showed that the watershed is at crucial water resources management turning point. The WRM-PCJ performed well, and it proved to be an excellent tool for decision and policy makers at RB-PCJ.     

Using Optimized Deep Learning to Predict Daily Streamflow: A Comparison to Common Machine Learning Algorithms

From a watershed management perspective, streamflow need to be predicted accurately using simple, reliable, and cost-effective tools. Present study demonstrates the first applications of a novel optimized deep-learning algorithm of a convolutional neural network (CNN) using BAT metaheuristic algorithm (i.e., CNN-BAT). Using the prediction powers of 4 well-known algorithms as benchmarks – multilayer perceptron (MLP-BAT), adaptive neuro-fuzzy inference system (ANFIS-BAT), support vector regression (SVR-BAT) and random forest (RF-BAT), the CNN-BAT model is tested for daily streamflow (Q_t) prediction in the Korkorsar catchment in northern Iran. Fifteen years of daily rainfall (R_t) and streamflow data from 1997 to 2012 were collected and used for model development and evaluation. The dataset was divided into two groups for building and testing models. The correlation coefficient (r) between rainfall and streamflow with and without antecedent events (i.e., R_t-1, R_t-2, etc.) (as the input variables) and Q_t (as the output variable) served as the basis for constructing different input scenarios. Several quantitative and visually-based evaluation metrics were used to validate and compare the model’s performance. The results indicate that R_t was the most effective input variable on Q_t prediction and the integration of R_t, R_t-1, and Q_t-1 was the optimal input combination. The evaluation metrics show that the CNN-BAT algorithm outperforms the other algorithms. The Friedman and Wilcoxon signed-rank test indicates that the prediction power of CNN-BAT algorithm is significantly/statistically different from the other developed algorithms.

     

Extreme water level rise across the upper Laurentian Great Lakes region: Citizen science documentation 2010–2020

As the global water balance accelerates in a warming climate, extreme fluctuations in the water levels of lakes and aquifers are anticipated, with biogeochemical, ecological and water supply consequences. However, it is unclear how site-specific factors, such as location, morphometry and hydrology, will modulate these impacts on regional spatial scales. Here, we report water level time series collected by citizen scientists for 15 diverse inland lakes in the upper Laurentian Great Lakes region from 2010 to 2020, and we compare these time series with those for the two largest Great Lakes, Lake Superior and Lake Michigan-Huron. Combined with historical data (1942–2010), the findings indicate that lakes spanning seven orders of magnitude in size (10^?2 to 10⁵ km²) all rebounded from record low to record high water levels during the recent decade. They suggest coherent water level oscillations among regional lakes (large and small) implying a common, near-decadal, climatic driver that may be changing.     

Daily prediction of marine beach water quality in Hong Kong

Bacterial concentration (Escherichia coli) is generally adopted as a key indicator of beach water quality. Currently the beach management system in Hong Kong relies on past water quality data sampled at intervals between 3 and 14 days. Beach advisories are issued when the geometric mean E. coli level of the past five samples exceeds the beach water quality objective (WQO) of 180 counts/100 mL. When the E. coli level varies dynamically, the system is not able to track the daily bacterial variation. And yet worldwide there does not exist a generally accepted method to predict beach water quality in a marine environment, which is influenced by hydro-meteorological variables, catchment characteristics, as well as complicated tidal currents and wave effects.A comprehensive study of beach water quality prediction has been carried out for four representative beaches in Hong Kong: Big Wave Bay (BW), Deep Water Bay (DW), New Cafeteria (NC) and Silvermine Bay (SIL). Statistical analysis of the extensive regular monitoring data was carried out for two periods before and after the commissioning of the Harbour Area Treatment Scheme (HATS): (1990–1997) and (2002–2006) respectively. The data analysis shows that E. coli is strongly correlated with seven hydro-environmental variables: rainfall, solar radiation, wind speed, tide level, salinity, water temperature and past E. coli concentration. The relative importance of the parameters is beach-specific, and depends on the local geographical and hydrographical characteristics as well as location of nearby pollution sources.Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models are developed from the sparsely sampled regular monitoring data (2002–2006) to predict the next-day E. coli concentration using the key hydro-environmental variables as input parameters. The models are validated against daily monitoring data in the bathing seasons of 2007 and 2008. The models are able to track the dynamic changes in E. coli concentration and predict WQO compliance/exceedance with an overall accuracy of 70–96%. Both the MLR and ANN models are superior to the current beach advisories in capturing water quality variations, and in predicting WQO exceedances. For example, the models predict around 80% and 50% of the exceedances at BW and NC respectively in June–July 2007, as compared to 0% and 14% based purely on past data. Similarly, observed exceedances are predicted with success rates of 71%, 42%, and 53% at BW, NC, and SIL respectively during July–October 2008, as compared with 0%, 0%, and 6% using the current water quality assessment criterion. The MLR and ANN models have similar performances; ANN model tends to be better in predicting the high-end concentrations, with however a greater number of false positive predictions (false alarms).This work demonstrates the practical feasibility of predicting bacterial concentration based on the critical hydro-environmental variables, and paves the way for developing a real time water quality forecast and management system for Hong Kong.     

A Wavelet-ANFIS Hybrid Model for Groundwater Level Forecasting for Different Prediction Periods

Artificial neural network (ANN) and Adaptive Neuro-Fuzzy Inference System (ANFIS) have an extensive range of applications in water resources management. Wavelet transformation as a preprocessing approach can improve the ability of a forecasting model by capturing useful information on various resolution levels. The objective of this research is to compare several data-driven models for forecasting groundwater level for different prediction periods. In this study, a number of model structures for Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS), Wavelet-ANN and Wavelet-ANFIS models have been compared to evaluate their performances to forecast groundwater level with 1, 2, 3 and 4 months ahead under two case studies in two sub-basins. It was demonstrated that wavelet transform can improve accuracy of groundwater level forecasting. It has been also shown that the forecasts made by Wavelet-ANFIS models are more accurate than those by ANN, ANFIS and Wavelet-ANN models. This study confirms that the optimum number of neurons in the hidden layer cannot be always determined by using a specific formula but trial-and-error method. The decomposition level in wavelet transform should be determined according to the periodicity and seasonality of data series. The prediction of these models is more accurate for 1 and 2 months ahead (for example RMSE?=?0.12, E?=?0.93 and R ²?=?0.99 for wavelet-ANFIS model for 1 month ahead) than for 3 and 4 months ahead (for example RMSE?=?2.07, E?=?0.63 and R ²?=?0.91 for wavelet-ANFIS model for 4 months ahead).     

Prediction and simulation of monthly groundwater levels by genetic programming

Groundwater level is an effective parameter in the determination of accuracy in groundwater modeling. Thus, application of simple tools to predict future groundwater levels and fill-in gaps in data sets are important issues in groundwater hydrology. Prediction and simulation are two approaches that use previous and previous-current data sets to complete time series. Artificial intelligence is a computing method that is capable to predict and simulate different system states without using complex relations. This paper investigates the capability of an adaptive neural fuzzy inference system (ANFIS) and genetic programming (GP) as two artificial intelligence tools to predict and simulate groundwater levels in three observation wells in the Karaj plain of Iran. Precipitation and evaporation from a surface water body and water levels in observation wells penetrating an aquifer system are used to fill-in gaps in data sets and estimate monthly groundwater level series. Results show that GP decreases the average value of root mean squared error (RMSE) as the error criterion for the observation wells in the training and testing data sets 8.35 and 11.33 percent, respectively, compared to the average of RMSE by ANFIS in prediction. Similarly, the average value of RMSE for different observation wells used in simulation improves the accuracy of prediction 9.89 and 8.40 percent in the training and testing data sets, respectively. These results indicate that the proposed prediction and simulation approach, based on GP, is an effective tool in determining groundwater levels.     

Application of Several Data-Driven Techniques for Predicting Groundwater Level

In this study, several data-driven techniques including system identification, time series, and adaptive neuro-fuzzy inference system (ANFIS) models were applied to predict groundwater level for different forecasting period. The results showed that ANFIS models out-perform both time series and system identification models. ANFIS model in which preprocessed data using fuzzy interface system is used as input for artificial neural network (ANN) can cope with non-linear nature of time series so it can perform better than others. It was also demonstrated that all above mentioned approaches could model groundwater level for 1 and 2 months ahead appropriately but for 3 months ahead the performance of the models was not satisfactory.     

Comparison of Two Different Adaptive Neuro-Fuzzy Inference Systems in Modelling Daily Reference Evapotranspiration

This study compares two different adaptive neuro-fuzzy inference systems, adaptive neuro-fuzzy inference system (ANFIS) with grid partition (GP) method and ANFIS with subtractive clustering (SC) method, in modeling daily reference evapotranspiration (ET ₀ ). Daily climatic data including air temperature, solar radiation, relative humidity and wind speed from Adana Station, Turkey were used as inputs to the fuzzy models to estimate daily ET ₀ values obtained using FAO 56 Penman Monteith (PM) method. In the first part of the study, the effect of each climatic variable on FAO 56 PM ET ₀ was investigated by using fuzzy models. Wind speed was found to be the most effective variable in modeling ET ₀ . In the second part of the study, the effect of missing data on training, validation and test accuracy of the neuro-fuzzy models was examined. It was found that the ANFIS-GP model was not affected by missing data while the test accuracy of the ANFIS-SC model slightly decreases by increasing missing data’s percent. In the third part of the study, the effect of training data length on training, validation and test accuracy of the ANFIS models was investigated. It was found that training data length did not significantly affect the accuracy of ANFIS models in modeling daily ET ₀ . ANFIS-SC model was found to be more sensitive to the training data length than the ANFIS-GP model. In the fourth part of the study, both ANFIS models were compared with the following empirical models and their calibrated versions; Valiantzas’ equations, Turc, Hargreaves and Ritchie. Comparison results indicated that the three-and four-input ANFIS models performed better than the corresponding empirical equations in modeling ET ₀ while the calibrated two-parameter Ritchie and Valiantzas’ equations were found to be better than the two-input ANFIS models.     

Predicting Shallow Water Table Depth at Regional Scale: Optimizing Monitoring Network in Space and Time

Shallow water table levels can be predicted using several approaches, either based on climatic records, on field evidences based on soil morphology, or on the outputs of physically based models. In this study, data from a monitoring network in a relevant agricultural area of Northern Italy (ca. 12,000 Km²) were used to develop a data driven model for predicting water table depth in space and time from meteorological data and long-term water table characteristics and to optimize sampling density in space and time. Evolutionary Polynomial Regressions (EPR) were used to calibrate a predictive tool based on climatic data and on the records from 48 selected sites (N?=?5,611). The model was validated against the water table depths observed in 15 independent sites (N?=?1,739), resulting in a mean absolute error of 30.8 cm (R ²?=?0.61). The model was applied to the whole study area, using the geostatistical estimates of the average water table depth as input, to provide spatio-temporal maps of the water table depth. The impact of the degradation of data input in the temporal and spatial domain was then assessed following two approaches. In the first case, three different EPR models were calibrated based on 25 %, 50 % and 75 % of the available data, and the error indexes compared. In the second case, an increasing number of monitoring sites were removed from the initial data set, and the associated increased kriging standard deviation was assessed. Reducing the average sampling frequency from 1.5 per month to 1 every 40 days did not impact significantly on the prediction capability of the proposed model. Reducing the sampling frequency to 1 every 4 months resulted in a loss of accuracy <3 %, while removing more than half locations from the network, resulted in a global loss of information <15 %.     

Variability of light absorption properties in optically complex inland waters of Lake Chaohu,China

Absorption coefficients of phytoplankton, colored detrital matter (CDM), non-algal particles (NAP), colored dissolved organic matter (CDOM), and their relative contributions to total non-water absorption (a_t ? w) are essential variables for bio-optical and radiative transfer models. Light absorption properties showed large range and variability sampled at 194 stations throughout Lake Chaohu between May 2013 and April 2015. The a_t ? w was dominated by phytoplankton absorption (a_ph) and NAP absorption (a_d). The contribution of CDOM absorption to a_t ? w was lower than 30%. Phytoplankton and NAP were the primary sources of spatial and vertical variability in absorption properties. Light absorption by CDOM, though significant in magnitude, was relatively constant. CDM absorption (a_dg) was dominated by NAP. The spatial variation of the absorption coefficients from each of the optically active constituents were driven by several main inflow rivers in the western and middle part of Lake Chaohu. Algal blooms and bottom resuspension contributed to vertical variability as observed by phytoplankton and NAP profiles. Specific absorption of phytoplankton had significant spatial and seasonal variations without vertical variation. The spectral slope of absorption showed no significant spatial variability (p > 0.05). Variations of absorption affected different ranges of remote sensing reflectance (R_rs) spectrum, thereby increasing the difficulty of applying the remote sensing algorithm in optically complex waters. Parameters and relationships presented in this study provide useful information for bio-optical models and remote sensing of lakes similar to Lake Chaohu in terms of optical properties.     

Sums, Products and Ratios for Crovelli’s Bivariate Gamma Distribution

Bivariate gamma distributions have been used successfully on modeling hydrological processes. In this work, supposing that X and Y follow the Crovelli’s bivariate gamma model, we deduce the exact distributions of the functions U?=?X?+?Y, P?=?XY and Q?=?X/(X?+?Y), as well as their respective moments. Those functions are important hidrological variables. A MAPLE code to compute the quantiles is provided. An application of the results is provided to rainfall data from Passo Fundo.     

Lake Water-Level fluctuations forecasting using Minimax Probability Machine Regression,Relevance Vector Machine,Gaussian Process Regression,and Extreme Learning Machine

Forecasting freshwater lake levels is vital information for water resource management, including water supply management, shoreline management, hydropower generation optimization, and flood management. This study presents a novel application of four advanced artificial intelligence models namely the Minimax Probability Machine Regression (MPMR), Relevance Vector Machine (RVM), Gaussian Process Regression (GPR) and Extreme Learning Machine (ELM) for forecasting lake level fluctuation in Lake Huron utilizing historical datasets. The MPMR is a probabilistic framework that employed Mercer Kernels to achieve nonlinear regression models. The GPR, which is a probabilistic technique used tractable Bayesian framework for generalization of multivariate distribution of input samples to vast dimensional space. The ELM is a capable algorithm-based model for the implementation of the single-layer feed-forward neural network. The RVM demonstrate depends on the specification of the Bayesian method on a linear model with proper preceding that results in demonstration of sparse. The recommended techniques were tested to evaluate the current lake water-level trend monthly from the historical datasets at four previous time steps. The Lake Huron levels from 1918 to 1993 was managed for the training phase, and the rest of data (from 1994 to 2013) was used for testing. Considering the monthly and annually previous time steps, six models were introduced and found that the best results are achieved for a model with (t-1, t-2, t-3, t-12) as input combinations. The results show that all models can forecast the lake levels precisely. The results of this research study exhibit that the MPMR model (R²?=?0.984; MAE?=?0.035; RMSE?=?0.044; E_NS?=?0.984; D_Refined?=?0.995; E_LM?=?0.874) found to be more precise in lake level forecasting. The MPMR can be utilized as a practical computational tool on current and future planning with sustainable management of water resource of Lake Michigan-Huron.

     

Warmer and Drier Climates that Make Terminal Great Lakes

A recent empirical model of glacial-isostatic uplift showed that the Huron and Michigan lake level fell tens of meters below the lowest possible outlet about 7,900 ¹⁴C years BP when the upper Great Lakes became dependent for water supply on precipitation alone, as at present. The upper Great Lakes thus appear to have been impacted by severe dry climate that may have also affected the lower Great Lakes. While continuing paleoclimate studies are corroborating and quantifying this impacting climate and other evidence of terminal lakes, the Great Lakes Environmental Research Laboratory applied their Advanced Hydrologic Prediction System, modified to use dynamic lake areas, to explore the deviations from present temperatures and precipitation that would force the Great Lakes to become terminal (closed), i.e., for water levels to fall below outlet sills. We modeled the present lakes with pre-development natural outlet and water flow conditions, but considered the upper and lower Great Lakes separately with no river connection, as in the early Holocene basin configuration. By using systematic shifts in precipitation, temperature, and humidity relative to the present base climate, we identified candidate climates that result in terminal lakes. The lakes would close in the order: Erie, Superior, Michigan-Huron, and Ontario for increasingly drier and warmer climates. For a temperature rise of T°C and a precipitation drop of P% relative to the present base climate, conditions for complete lake closure range from 4.7T + P > 51 for Erie to 3.5T + P > 71 for Ontario.     

Waxing and waning slacks: The changing ecohydrology of interdunal wetlands/slacks in a Lake Michigan coastal dune complex during rising Lake Michigan-Huron levels

Since 2016 we have studied the largest interdunal wetlands/slack lying within a deflated parabolic dune east of Lake Michigan. Geologic cross-sections show ∼ 15 m of sand and gravel beneath the dunes, creating an aquifer hydraulically connecting Lake Michigan-Huron (MH) with the water table/shallow groundwater influencing the slack. Lake Michigan-Huron (MH) water levels have risen ∼ 1 m from 2016 to 2020, increasing water levels within and around the slack ∼ 1 m. Color-infrared images and vegetation quadrat sampling show water appearing, then significantly expanding with the main slack and upland/dune vegetation transitioning to wetland vegetation in response to this rise. Monitoring well data show slack water levels rise in spring as Lake MH rises. Levels drop as the growing season begins while Lake MH continues to rise through summer. Short-term slack water level increases occur due to local rain events, but significant water level declines follow due to evapotranspiration. Slack water levels begin to rise again in late summer and into fall as the end of the growing season arrives, evapotranspiration decreases, and heavier, more frequent rain events occur. Together, these factors push slack water levels to their highest point of the year while Lake MH levels are decreasing. In late fall–winter, slack water levels drop in concert with Lake MH levels. Climate change effects, increased transpiration from higher temperatures, summer drought, and greater variability in lake level fluctuations, may make it more difficult to maintain wet growing conditions for hydrophytic vegetation. Hence, climate change poses risks to the existence of this imperiled ecosystem.     

Simulation of energy dissipation downstream of labyrinth weirs by applying support vector regression integrated with meta-heuristic algorithms

In this study, multi-tracker optimization algorithm (MTOA), particle swarm optimization (PSO), and differential evolution (DE) algorithms were integrated with support vector regression (SVR) to predict energy dissipation downstream of labyrinth weirs (ΔE). In order to evaluate the performance of these methods, the results are compared with corresponding outcome obtained by applying two other methods, namely, multilayer perceptron neural network (MLPNN) and multiple linear regressions methods (MLR). The input parameters comprise the discharge, the upstream flow depth, the crest length of a single cycle of the labyrinth weir, the width of a single cycle of the labyrinth weir, the apex width, the number of labyrinth weir cycles, the sidewall angle, and the height of weir. The results indicate that the meta-heuristic algorithms substantially improve the performance of SVR. The results show that the integrative methods, SVR-MTOA, SVR-PSO, and SVR-DE, are more accurate than the MLPNN and the MLR. In average, the integrative methods provide 39.63% more accurate results than the MLPNN and 79.34% more accurate results than the MLR. The average RMSE and R² for the integrative methods are 0.0054 m and 0.977, respectively. Among all integrative methods, the SVR-MTOA yields the best results, with RMSE = 0.0044 m and R² = 0.986.     

GIS Integrated Modelling of a Himalayan Landscape

Land-surface modelling is traditionally based on reductionistic cause-effect models developed for the temperate region. Sustainable management of vulnerable and extreme regions demands a new holistic approach relying on first principles and integration of processes and patterns. In this article remotely sensed data and GIS are combined for creating digital data sets of elevation and vegetation over the Himalayan Sutlej river and its tributaries. GIS-coupled models are used for distributed estimates of precipitation,and modelling of the basin water cycle. Based on the derived data and their scale and error, an expert system incorporating fuzzy logic is used for index-related erosion modelling. It is concluded that GIS integrated modelling can pave the way to sustainable landscape management.     

Experimental and numerical analysis of flow over a rectangular full-width sharp-crested weir

Weirs are a type of hydraulic structure, used for water level adjustment, flow measurement, and diversion of water in irrigation systems. In this study, experiments were conducted on sharp-crested weirs under free-flow conditions and an optimization method was used to determine the best form of the discharge coefficient equation based on the coefficient of determination(R~2) and root mean square error(RMSE). The ability of the numerical method to simulate the flow over the weir was also investigated using Fluent software. Results showed that, with an increase of the ratio of the head over the weir crest to the weir height(h/P), the discharge coefficient decreased nonlinearly and reached a constant value of 0.7 for h/P 0.6. The best form of the discharge coefficient equation predicted the discharge coefficient well and percent errors were within a ±5%error limit. Numerical results of the discharge coefficient showed strong agreement with the experimental data. Variation of the discharge coefficient with Reynolds numbers showed that the discharge coefficient reached a constant value of 0.7 when h/P 0.6 and Re 20000.     

Real-Time Water-Level Forecasting Using Dilated Causal Convolutional Neural Networks

Accurate forecasts of hourly water levels during typhoons are crucial to disaster emergency response. To mitigate flood damage, the development of a water-level forecasting model has played an essential role. We propose a model based on a dilated causal convolutional neural network (DCCNN) that can yield water-level forecasts with lead times of 1- to 6-h. A DCCNN model can efficiently exploit a broad-range history. Residual and skip connections are also applied throughout the network to enable training of deeper networks and to accelerate convergence. To demonstrate the superiority of the proposed forecasting technique, we applied it to a dataset of 16 typhoon events that occurred during the years 2012–2017 in the Yilan River basin in Taiwan. In order to examine the efficiency of the improved methodology, we also compared the proposed model with two existing models that were based on the multilayer perceptron (MLP) and the support vector machine (SVM). The results indicate that a DCCNN-based model is superior to both the SVM and MLP models, especially for modeling peak water levels. Much of the performance improvement of the proposed model is due to its ability to provide water-level forecasts with a long lead time. The proposed model is expected to be particularly useful in support of disaster warning systems.

     

Artificial Neural Network (ANN) Based Modeling for Karstic Groundwater Level Simulation

A relatively new method of addressing different hydrological problems is the use of artificial neural networks (ANN). In groundwater management ANNs are usually used to predict the hydraulic head at a well location. ANNs can prove to be very useful because, unlike numerical groundwater models, they are very easy to implement in karstic regions without the need of explicit knowledge of the exact flow conduit geometry and they avoid the creation of extremely complex models in the rare cases when all the necessary information is available. With hydrological parameters like rainfall and temperature, as well as with hydrogeological parameters like pumping rates from nearby wells as input, the ANN applies a black box approach and yields the simulated hydraulic head. During the calibration process the network is trained using a set of available field data and its performance is evaluated with a different set. Available measured data from Edward??s aquifer in Texas, USA are used in this work to train and evaluate the proposed ANN. The Edwards Aquifer is a unique groundwater system and one of the most prolific artesian aquifers in the world. The present work focuses on simulation of hydraulic head change at an observation well in the area. The adopted ANN is a classic fully connected multilayer perceptron, with two hidden layers. All input parameters are directly or indirectly connected to the aquatic equilibrium and the ANN is treated as a sophisticated analogue to empirical models of the past. A correlation analysis of the measured data is used to determine the time lag between the current day and the day used for input of the measured rainfall levels. After the calibration process the testing data were used in order to check the ability of the ANN to interpolate or extrapolate in other regions, not used in the training procedure. The results show that there is a need for exact knowledge of pumping from each well in karstic aquifers as it is difficult to simulate the sudden drops and rises, which in this case can be more than 6 ft (approx. 2 m). That aside, the ANN is still a useful way to simulate karstic aquifers that are difficult to be simulated by numerical groundwater models.