You are on page 1of 1

ASSESSMENT OF THE UNCERTAINTIES AND

COMPLEXITY IN THE CONCEPTUAL MODEL IN


NUMERICAL MODELLING OF SALINE
INTERFACE.
C. Canul-Macario*, P. Salles*,+, J. A. Hernández-Espriú** and R. Pacheco Castro*,+
* Engineering and Coastal Process Laboratory, Sisal, Yucatan, Mexico. Engineering Institute, UNAM (e-mail:
CCanulM@iingen.unam.mx; PSallesA@iingen.unam.mx; RPachecoC@iingen.unam.mx)
+ National Coastal Resilience Laboratory (www.lanresc.mx)

** Hydrogeolgy Group, Engineering Faculty, UNAM (E-mail: ahespriu@unam.mx)

1. INTRODUCTION 4. NUMERICAL MODEL IMPLEMENTATION


Coastal zones are important development regions in the world and aquifers are often the
main supply for freshwater in coastal populations. Numerical modelling is a tool for
improving management strategies in coastal aquifers (Ketabchi et al, 2016; Werner et al.,
2013). Numerical modelling of aquifers poses a scientific challenge, due to uncertainties in
the conceptual model. In these cases, evaluating different conceptual models is a strategy
that can improve the calibration and results of numerical models (Gondwe et al. 2011;
Refsgaard et al. 2012). This work shows a numerical model of flow and saline interface in
the karstic aquifer of northwest Yucatan (Mexico) and discusses how the model was chosen
based on the evaluation of different conceptual models.

2. MATERIALS AND METHODS


Hydrogeologic Characterization Figure 3. Conceptual models tested. Figure 4. Numerical model implementation
Literature + Field Data
PEST = Parameter estimation A 2D numerical model of flow and transport (600 columns and 27 layers) were implemented using the
optimization. SEAWAT 2000 (Guo & Langevin, 2002) for a coastal transect (Fig. 4). Time series of 10 days of hydraulic
Aq. Conceptual Models: M1-Homegeneous, M2- RMSE = Root mean squared heads and salinity profiles were used to simulate: (a) the groundwater flow, (b) the off-shore tide propagation
Lateral Heterogenous, M3-Vertical Heterogeneous, error. into the aquifer and (c) the freshwater–saltwater interface. Model is inspired by anisotropic dispersive Henry
M4-Vertical Lateral Heterogeneous. r = Pearson correlation. problem (Abarca et al. 2007). The models were validated in drought and rainy season.
NSE = Nash-Sutcliffe
efficiency.
Modelling SEAWAT (VMODFLOW) AIC = Akaike information 5. RESULTS AND DISCUSSION
Conceptual Models + Field Data criteria. Table 1 shows that the M4 model calibrates better, but for validation, this model performance is worse when
wjAIC = Akaike weights compared to less complex numerical models. This means that the most complex model, M4, is not the best
BIC = Bayesian information choice for the simulation of the hydraulic head in the aquifer. For saline interface position, M2 is the numerical
Groundwater Numerical Model Calibration- criteria. model with the smallest error during the calibration, but the results also suggest that M3 can be an option to
Validation and Election simulate the saline interface. The validation shows that M3 is the best numerical model for the hydraulic head
(Hill and Tiedeman 2007) and saline interface simulation (Table 1). Therefore, we chose M3 as the best conceptual model for the
PEST + RMSE + r + NSE (Accuracy)
simulation of flow and transport in the study zone. This exercise shows that the most complex models can
AIC + wjAIC + BIC (Complexity) calibrate better than less complex models, but it does not imply that it can better simulate the behaviour of the
Figure 1. Methodology flux diagram. aquifer under different conditions (changes in the boundaries and different time discretization, specifically).
Calibration with tide effects
time
Hydraulic Head Salinity propagation Aquifer

3. FIELD SITE ID
RMSE RMSE
error (hrs) discharge
(m3/s/km)
r NSE AIC wjAIC BIC r NSE AIC wjAIC BIC Min Max
(m) (g/l)
M1 0.075 0.93 0.84 -2333 0 -5351 3.27 0.94 0.94 353.5 0.01 836 0 2 0.075
M2 0.094 0.93 0.74 -2115 0 -4843 2.27 0.97 0.96 346.2 0.49 820 0 1 0.273
M3 0.086 0.94 0.78 -2196 0 -5030 2.79 0.96 0.96 346.5 0.42 820 0 1 0.25
M4 0.054 0.96 0.91 -2603 1 -5952 3.02 0.96 0.95 350.0 0.07 828 0 2.5 0.15
Validation discharge season
Table 1. Results of
M1 0.116 0.83 0.31 -443 0.00 -1010 5.06 0.92 0.87 357.6 0 845
calibration and
M2 0.074 0.85 0.72 -5320 0.01 -1213 4.85 0.92 0.88 355.9 0 842 validation of the
M3 0.070 0.86 0.74 -541 0.99 -1234 2.61 0.96 0.96 330.0 1 782 multiple
M4 0.124 0.88 0.2 -408
0.00 -925 4.93 0.92 0.88 356.5 0 843 conceptual
Validation recharge season models. Blue
M1 0.157 0.88 -0.07 -377.8 0.00 -859 4.08 0.94 0.91 348.7 0 825 values show the
M2 0.077 0.92 0.59 -523.1 0.46 -1192 3.87 0.94 0.92 346.4 0 820 best fit, red values
M3 0.077 0.90 0.59 -523.4 0.54 -1193 2.45 0.96 0.96 327.5 1 776 show a rejected
M4 0.172 0.91 -1.07 -339.0 0.00 -766 4.29 0.93 0.90 350.8 0 830 criterion.

Figure 2. Study Zone


The NW Coastal Aquifer of Yucatán (RNWY) is located at SW of Mexican Republic in a
Tertiary and Quaternary rock (Fig. 2). The study zone is defined as a confined coastal Figure 5. Results of
simulation.
karstic aquifer. Regional groundwater flow is perpendicular to the coast with a low (a) Hydraulic head,
hydraulic gradient 1X10-5 m/m (Villasuso et al., 2011). The aquifer characterization (b) Salinity
evidences the existence of heterogeneities, thus the need to assess multiple conceptual
models: (M1) homogeneous aquifer, (M2) heterogenous-lateral aquifer, (M3)
heterogeneous-vertical aquifer and (M4) heterogeneous-vertical-lateral aquifer (Fig. 3).

6. CONCLUSIONS
It is important to assess not only the accuracy of the model but also the complexity of the conceptual model. In this direction, it may be preferable to select the model with fewer parameters, due to
generally it is complicated to validate the most complex models. The calibration is not a definitive criterion for the selection of the numerical model setup, because increasing the parameters can lead to the
best calibration; but later, show poor performance to validation.
The numerical model results show the thickness of the freshwater aquifer, the thickness of the salt interface and the distribution of salts throughout the domain. The salinity results are important to
understand the spatial-temporal variation of the salinity in the groundwater of the coastal ecosystems. This numerical model is the first step to evaluate climate change scenarios and their effects in the
aquifer head, groundwater salinity and the saline interface position in RNWY aquifer.

7. REFERENCES
Abarca, Elena, Jesús Carrera, Xavier Sánchez-Vila, and Marco Dentz. 2007. “Anisotropic Dispersive Henry Problem.” Advances in Water Resources 30 (4):913–26. https://doi.org/10.1016/j.advwatres.2006.08.005.
Gondwe, Bibi R N, Gonzalo Merediz-Alonso, and Peter Bauer-Gottwein. 2011. “The Influence of Conceptual Model Uncertainty on Management Decisions for a Groundwater-Dependent Ecosystem in Karst.” Journal of Hydrology 400 (1–2). Elsevier B.V.:24–40. https://doi.org/10.1016/j.jhydrol.2011.01.023.
Guo, Weixing, and Christian D. Langevin. 2002. User’s Guide to SEAWAT: A Computer Program for Simulation of Three-Dimensional Variable-Density Ground-Water Flow. USGS Techniques of Water Resources Investigations. Miami Florida: USGS.
Ketabchi, Hamed, Davood Mahmoodzadeh, Behzad Ataie-Ashtiani, and Craig T. Simmons. 2016. “Sea-Level Rise Impacts on Seawater Intrusion in Coastal Aquifers: Review and Integration.” Journal of Hydrology 535. Elsevier B.V.:235–55. https://doi.org/10.1016/j.jhydrol.2016.01.083.
Refsgaard, Jens Christian, Steen Christensen, Torben O. Sonnenborg, Dorte Seifert, Anker Lajer Højberg, and Lars Troldborg. 2012. “Review of Strategies for Handling Geological Uncertainty in Groundwater Flow and Transport Modeling.” Advances in Water Resources 36: 36–50. https://doi.org/10.1016/j.advwatres.2011.04.006.
Hill, Mary C., and Claire R. Tiedeman. 2007. “Evaluating Model Fit.” In Effective Groundwater Model Calibration, 93–123. https://doi.org/10.1002/9780470041086.ch6.
Villasuso-Pino, M., I. Sanchez y Pinto, C. Canul Macario, G. Baldazo Escobedo, R. Casares Salazar, J. Souza Cetina, P. Poot Euan, and C. Pech. 2011. “Hydrogeology And Conceptual Model Of The Karstic Coastal Aquifer In Northern Yucatan State, Mexico.” Tropical and SubtroTropical Agroecosystems 13:243–60.
Werner, Adrian D., Mark Bakker, Vincent E.A. Post, Alexander Vandenbohede, Chunhui Lu, Behzad Ataie-Ashtiani, Craig T. Simmons, and D. A. Barry. 2013. “Seawater Intrusion Processes, Investigation and Management: Recent Advances and Future Challenges.” Advances in Water Resources 51. Elsevier Ltd:3–26.
https://doi.org/10.1016/j.advwatres.2012.03.004.

This project has received funding from the National Coastal Resilience Laboratory (LANRESC) and
CONACYT.

You might also like