HydroPredict 2008 – Prague, Czech Republic; Bruthans-Kovar-Hrkal (eds.)
357
Coupling monitoring networks and regional scale flow models for the management of groundwater resources: the Almádena-Odeáxere Aquifer case study (Algarve-Portugal) J. MARTINS & J. P. MONTEIRO Algarve University Geo-Systems Centre UALG/CVRM, Marine and Environmental Sciences Faculty, Campus de Gambelas 8005-139 Faro, Portugal
[email protected]
Abstract Automatic monitoring networks were settled to provide datasets for the calibration and validation of regional flow models, as a strategy to complement data available in official monitoring networks for the main regional aquifers in Algarve (Portugal). The AlmádenaOdeáxere aquifer system is presented as an example, where the comparative analysis of variables representing “reality”, consisting of data obtained in these networks, was made against values simulated with a finite element numeric model. This methodology contributed for a better calibration and validation of the model, together with the design of effective groundwater monitoring networks, at the regional scale. Keywords inverse calibration; use of models to guide data collection; use of data to calibrate models; Algarve
INTRODUCTION The 17 aquifer systems presently identified in the Algarve region (South Portugal) are currently being monitored by the Coordinating Commission of the Algarve - CCDR (official entity) through a set of monitoring networks, comprising 138 observation points, which control several state variables, relevant for water management. However, the density of these networks and the respective frequency of data collection cannot provide datasets detailed enough to characterise the spatial distribution and temporal evolution of these variables in order to allow the calibration and validation of quantitative regional flow models. In order to overcome these difficulties on the Almádena-Odeáxere aquifer system (AO), the work undertaken consisted in coupling: (1) a regional finite element flow model, developed to investigate the hydraulic behaviour of the AO; (2) the setting up of an automatic monitoring network of 10 additional in-situ environmental sensors for real-time monitoring, custom-designed in a complementary way the official monitoring network already existing at AO (11 observation points). Subsequent adjustments on the position of the settled observation points were made by continuously retrieving and analysing available head data. These adjustments were performed alongside with the purpose of calibrating the finite element flow model through inverse methods, thus providing the necessary model input data to fulfil this objective. MONITORING NETWORKS AND CONCEPTUAL FLOW MODEL The AO (Fig. 1) has an area of 63,5 km2 spanning from Odeáxere (East) to Almádena (West). It develops in carbonate Lias-Dogger lithologies (limestones, dolomitic limestones and dolostones), which show, in some places, a well developed karst with a thickness in the order of 750 m (Reis, 1993; Almeida et al., 2000). According to the conceptual model, regional water is expected to flow
358
predominantly from NE to SW (towards the effluent reach of “ribeira de Vale do Barão” creek) and N to S (towards the effluent reach of “ribeira de Bensafrim” creek).
Fig. 1 Hydraulic head contours, interpolated from the official monitoring network (“x” marks) data (dashed lines) and interpolated from the joint use of the implemented automatic monitoring network (circle symbols), represented as solid lines.
An analysis of the piezometric surface of AO (contoured using data obtained at the official monitoring network, from March 1978 to February 2007) showed that the available data was insufficient to provide a consistent estimate of the hydraulic behaviour of the aquifer. According to this data, flow predominantly occurred NW to SE, which should not happen in reality due to the existence of an impermeable geologic boundary at the SE limit of AO. The existent monitoring network lacked the necessary coverage, on important sectors of the aquifer, for an efficient calibration of the aquifer’s finite-element model. In order to complete the existing data, an additional automatic monitoring network was implemented at AO, in articulation with the process of calibrating the regional groundwater flow model, built for this aquifer. The elaboration of potentiometric surface maps based on the blending of data from these two monitoring sources has contributed considerably for the subsequent definition of 16 constant transmissivity, T, zones inside AO (Fig. 2, right). The subdivision of these zones took place on the basis of the character of the piezometric contours because there is little obvious variation in geology throughout the study area. This assumption follows the methodology carried out by Doherty (1998) which has already led to good calibration results. A similar procedure has also been used by Monteiro et al. (2006), Monteiro et al. (2007) to study Algarve aquifer systems. In order to overcome the lack of information on hydraulic head values at the defined zones, which were necessary for a successful calibration, 22 auxiliary “fictional” data points were distributed throughout the aquifer area. Hydraulic head values assigned to these points were based on the interpolation results of real head data at each of the corresponding coordinates. The introduction of these points did not result on a change on the flow pattern revealed by piezometry data obtained on the monitoring networks (as can be observed on Fig. 2, left).
HydroPredict 2008 – Prague, Czech Republic; Bruthans-Kovar-Hrkal (eds.)
359
Fig. 2 On the left, contours of hydraulic heads based on two datasets: the dashed lines represent data from the official network (“x” marks on Fig. 1) blended with data extracted from the automatic monitoring network (lozenges), the solid lines were created by adding auxiliary fictional hydraulic head values (crosses) to this data. On the right, 16 constant T zones (limited by grey thick lines) were defined according to the character of the obtained potentiometric surface (solid lines, on the left and right).
Simulated hydraulic head, in meters above sea level
INVERSE CALIBRATION OF THE MODEL
593/5 23 Fict22 22 Fict20 21 Fict21 20 19 18 Fict18 17 Fict19 Fict16 16 15 Fict17 14 13 Fict14 12 11 Fict15 10 9 AO-02 Fict12 8 Fict11 602/242 AO-06 7 603/38 AO-08 Fict10 Fict13 6 Fict9 602/187 5 AO-10 AO-14 AO-16 4 Fict7 AO-01 Fict6 Fict8 3 Fict4 Fict5 2 Fict1Fict3 1 0 Fict2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
Hydraulic head computed from measurements, in meters above sea level
Fig. 3 Simulated (dashed line) and measured (solid line) hydraulic head contours (above). Scatter plot of simulated vs. measured hydraulic heads (left and below). Spatial distribution of T values along the 16 predefined zones, obtained using inverse calibration (right and below).
In the last two decades, hydraulic parameters were obtained for AO from pumping tests on individual boreholes. These values ranged from 25 to 8784 m2 day-1 (Reis, 1993; Almeida et al., 2000). However these methods cannot provide the necessary data to carry out realistic representations of aquifers at the regional scale.
360
The performed calibration consisted consequently on the first attempt to obtain regionally distributed values of T using a synthetic bi-dimensional numerical representation of the AO (Martins, 2007). This hydraulic parameter was estimated by inverse modelling, using the Gauss-Marquardt-Levenberg method, implemented in the nonlinear parameter estimation software PEST (Doherty, 2002). The calibration results (Fig. 3), obtained using the inverse method, ranged from 86 m2 day-1 to 8158 m2 day-1 on 16 predefined zones of equal transmissivity. The calibration revealed a good fit between simulated and measured head values (Fig. 4), the correlation coefficient, R, value was higher than 0,9 (0,9967) and the sum-ofsquared weighted residuals between model outcomes and corresponding field data (i.e. the objective function, Φ) was 4,56 m. CONCLUSIONS Networks which automatically monitor hydraulic data in AO were designed in conjunction with the process of calibrating a numerical model through inverse methods. This model consists of a “synthetic representation” of the current knowledge regarding the hydrogeology of this aquifer. The Gauss-Marquardt-Levenberg method, implemented in the nonlinear parameter estimation software PEST, was used. Calibration results provided the first estimate on the aquifer’s regional transmissivity values distribution, which ranged from 86 m2 day-1 to 8158 m2 day-1. A good fit between simulated and measured head values was obtained, since the correlation coefficient, R, value was 0,9967 and the objective function, Φ, was 4,56 m. This work has contributed for improving the reliability of future simulations of spatial distribution and temporal evolution of state variables (hydraulic head and natural outflows), on natural conditions and also for different scenarios of water use. Acknowledgements The authors gratefully acknowledge the Portuguese Science and Technology Foundation (FCT) for the financial support of the POCTI/AMB/57432/2004 Project “Groundwater flow modelling and optimisation of groundwater monitoring networks at the regional scale in coastal aquifers – The Algarve case study” (CVRM Geo-Systems Centre, Algarve University). REFERENCES Almeida, C., Mendonça, J. L., Jesus, M. R. & Gomes, A. J. (2000) Sistemas Aquíferos de Portugal Continental. [Aquifer Systems of Mainland Portugal]. INAG. Relatório Técnico [Technical Report], CD-ROM, 661pp. Doherty, J. (1998) Use of MODFLOW in Groundwater Management in an Area of High Seasonal Rainfall. In: Proceedings of MODFLOW’98. Golden CO. U.S.A., 8pp. Doherty, J. (2002) PEST, Model-Independent Parameter Estimation. 4th Edition, Watermark Numerical Computing, Australia, 279 pp. Martins, J. (2007) Inverse Calibration of a Groundwater Flow Model for the Almádena-Odeáxere Aquifer System (Algarve – Portugal). MSc. Thesis, Faculdade de Ciências do Mar e do Ambiente, Universidade do Algarve. Monteiro, J. P., Vieira, J., Nunes, L. & Younes, F. (2006) Inverse Calibration of a Regional Flow Model for the QuerençaSilves Aquifer System (Algarve-Portugal). In: Integrated Water Resources Management and Challenges of the Sustainable Development. Marrakech. Morocco, CD-ROM, 6pp. Monteiro, J. P.; Oliveira, M. M. & Costa, J. P. (2007) Impact of the Replacement of Groundwater by Dam Waters in the Albufeira-Ribeira de Quarteira and Quarteira Coastal Aquifers. In: XXXV IAH Congress. Groundwater and Ecosystems. Lisbon. Portugal, pp 489-480, CD-ROM, 10pp. Reis, E. (1993) Estudo Hidrogeológico das Formações do Lias-Dogger situadas a ocidente do Rio Arade (Algarve), [Hydrogeologic Study of the Lias-Dogger Formations West of River Arade (Algarve)]. MSc. Thesis, FCUL.