Journal of Hydrology 563 (2018) 155–166

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Research papers

An upscaling procedure for the optimal implementation of open-loop T

geothermal energy systems into hydrogeological models
⁎
Sylvia Muela Mayaa, Alejandro García-Gilb,d, , Eduardo Garrido Schneiderb,d,
Miguel Mejías Morenoc, Jannis Eptinge, Enric Vázquez-Suñéf, Miguel Ángel Marazuelaf,
José Ángel Sánchez-Navarroa,d
a
Department of Earth Sciences, University of Zaragoza (UZ), c/ Pedro Cerbuna 12, 50009 Zaragoza, Spain
b
Geological Survey of Spain (IGME), C/ Manuel Lasala n° 44, 9° B, 50006 Zaragoza, Spain
c
Geological Survey of Spain (IGME) Headquarters, C/ Ríos Rosas 23, 28003 Madrid, Spain
d
Associated Unit in Earth Sciences IGME-UZ, Spain
e
Department of Environmental Sciences, Applied and Environmental Geology, University of Basel, Bernoullistr. 32, 4056 Basel, Switzerland
f
GHS, Institute of Environmental Assessment & Water Research (IDAEA), CSIC, Jordi Girona 18-26, 08034 Barcelona, Spain

A R T I C LE I N FO A B S T R A C T

This manuscript was handled by G. Syme, Different aspects of management policies for shallow geothermal systems are currently under development.
Editor-in-Chief, with the assistance of Jesús Although this technology has been used for a long time, doubts and concerns have been raised in the last years
Mateo-Lázaro, Associate Editor due to the massive implementation of new systems. To assess possible environmental impacts and manage
Keywords: subsurface energy resources, collecting data from operating shallow geothermal systems is becoming mandatory
Shallow geothermal energy in Europe. This study presents novel advances in the upscaling of operation datasets obtained from open-loop
GWHP geothermal energy systems for an optimal integration in hydrogeological models. The proposed procedure al-
Urban hydrogeology lows efficient numerical simulations to be performed at an urban scale. Specifically, this work proposes a novel
Thermal management
methodology to optimize the data treatment of highly transient real exploitation regimes by integrating energy
Groundwater
Thermal impact
transfer in the environment to reduce more than 90% registered raw datasets. The proposed methodology is then
applied to and validated on five different real optimization scenarios in which upscaling transformation of the
injection temperature series of 15-min sampling frequency has been considered. The error derived from each
approach was evaluated and compared for validation purposes. The results obtained from the upscaling pro-
cedures have proven the usefulness and transferability of the proposed method for achieving daily time functions
to efficiently reproduce the exploitation regimes of these systems with an acceptable error in a sustainable
resource management framework.

1. Introduction anthropogenic influences (Epting and Huggenberger, 2013). Heat ex-

change with the ground can be performed by different types of ground
Shallow geothermal systems are based on obtaining the heat energy source energy (GSE) systems, including closed or open systems, by
from materials of the most superficial layers (< 250 m) of the Earth́ s using heat pumps coupled with heat exchangers. Open systems, also
crust and the water that flows through them. The heat transfer from the called groundwater heat pumps (GWHPs), take direct advantage of the
Earth’s core to the outer areas of the crust and the capacity of the heat or cold of pumped groundwater and subsequently reinject pumped
ground to dampen thermal oscillations occurring on the surface make water into the aquifer (García-Gil et al., 2014a). GWHPs, and GSE in
thermal stability possible, starting at a depth of approximately 15 m. general, are widely used worldwide and their demand is expected to
After the damping of thermal oscillations with depth, the ground tem- increase in the next years (Epting et al., 2017; Jaudin, 2013; Lund and
perature is similar to the annual average temperature of the region plus Boyd, 2016). The increasing trend of GWHPs has resulted in an addi-
1 °C or 2 °C (Parsons, 1970). This terrain feature justifies the develop- tional heat load in urban aquifers, caused both by thermal interference
ment of these important systems as an adaptative measure to climate between systems and interference between well doublets, that is, the
change for renewable energy development (Bayer et al., 2012). The thermal autointerference effect (Galgaro and Cultrera, 2013; Garrido
‘potential natural state of the aquifer’ is defined as a state without et al., 2010b). Because the geothermal exploitation of the aquifer is not

⁎
Corresponding author at: Geological Survey of Spain (IGME), C/ Manuel Lasala n° 44, 9° B, 50006 Zaragoza, Spain.
E-mail address: [email protected] (A. García-Gil).

https://doi.org/10.1016/j.jhydrol.2018.05.057
Received 30 January 2018; Received in revised form 15 April 2018; Accepted 23 May 2018
Available online 26 May 2018
0022-1694/ © 2018 Elsevier B.V. All rights reserved.
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 1. Study area covering the metropolitan area of Zaragoza in the central sector of the Ebro River Basin (Spain), the confluence zone of the Gállego and Huerva
River tributaries.

consumptive, thermal pollution derived from these systems is the main therefore, the aquifer response to intensive shallow geothermal ex-
impact on the aquifer (Lo Russo et al., 2014). The problems associated ploitation, a requirement in obtaining a global vision necessary for
to the implementation of GWHPs originate both from the lack of energy aquifer management. Accomplishing this objective will help to under-
sustainability of the facilities due to thermal interference events stand the hydrodynamics and existing heat transport processes beneath
(Garrido et al., 2016) and from an insufficient legal framework. The urban environments, thus contributing to the improvement of sustain-
required regulatory frameworks to ensure sustainable exploitation of able management of shallow geothermal resources (Epting et al., 2013;
aquifer heat energy resources, particularly beneath cities where there Spitler, 2005). To reach this goal, different upscaling techniques were
may be overlapping and conflicting demands on the resource, are still considered in this work, to transform high-resolution datasets obtained
being developed, therefore generating great uncertainty for users from high frequency data logging into lower frequency data subsets
(García-Gil et al., 2015). (Bierkens et al., 2000; Finke et al., 2002). The upscaling procedure has
One of the possible ways to enhance the management of geothermal been widely used and developed in applied research typical in en-
systems is using numerical models to simulate the thermal regime in the vironmental science, where specific questions raised by society’s deci-
urban aquifer and to establish effective management strategies (Banks, sion makers (policy scale) and observation scale are not met. The scale
2009). This approach allows considering the high complexity of these transfer procedure or upscaling has been classified by Bierkens et al.
facilities, the heterogeneity of the medium, and the temporal variability (2000) depending on the involvement or non-involvement of a model in
of their operations in an integrated way. On the other hand, these the research cycle, the possible linear relationship between the model
methods have two limitations, that is, the high volume of data and the and input variables and parameters, the applicability of the model to
time requirements (García-Gil et al., 2014a). The modelling process of different locations/time steps, the form of the model at different scales
GWHP systems was carried out using data from specific discrete mea- or the possibility of deriving analytically a different scale model. The
surements in previous studies (Epting et al., 2013; Gropius, 2010; major classes of upscaling methods consist of averaging the observa-
Herbert et al., 2013). This simplification was applied as an appropriate tions or output variables (Brus and de Gruijter, 1997; Viscarra Rossel
approximation considering the inherent difficulties in the monitoring of et al., 2016), finding representative parameters or input variables
such complex installations and the consequent lack of exploitation re- (Dagan, 1981; Wu et al., 2006), averaging the model equations
gime datasets. Currently, regulators are increasingly requiring mon- (Bedrikovetsky, 2008; Whitaker, 1986) and performing a model sim-
itoring data on the operation of GWHP systems. These systems typically plification (de Vries et al., 1998; Vogler et al., 2018). The upscaling
operate following a design power, and thus, as building demand varies, procedure applied in this paper corresponds to the methods finding
the schemes may operate continuously or rather intermittently. Parti- representative parameters since the input variables of the model in-
cularly, the system may have a rather short operational cycling period volved are non-linear, so it cannot be applied at many time steps and
at locations where a scheme delivers a high proportion of the total the model has the same form at the two scales involved (Bierkens et al.,
heating or cooling demand. This leads to the need for very high fre- 2000). Furthermore, since it is not possible to obtain output variables at
quency monitoring data to characterise the system’s operation. In the larger scale and input data at the source scale is exhaustive, de-
contrast, numerical modelling becomes computationally expensive and terministic methods have been applied in accordance with previous
time consuming if this short cycling detail is to be represented ex- studies (Bierkens and van der Gaast, 1998; Johannes Dolman and Blyth,
plicitly. 1997; Yu et al., 2016). Finally, a sensitivity analysis was performed by
The main purpose of this study is to develop and validate a meth- numerically modelling different upscale scenarios to optimize the pro-
odology to obtain, as optimally and efficiently as possible, a subset of cedure and to perform and conduct an error assessment.
maximal representative data, which can be easily implemented in nu-
merical heat transport models, that is, to obtain data subsets for nu-
2. Study site
merical models resulting in minimum deviations when compared with
original data. Validation of the proposed methodology under standard
The present work was carried out within the framework of the urban
hydrogeological parameters is carried out to ensure its transferability to
alluvial aquifer of Zaragoza (Fig. 1) located in the central sector of the
other aquifers operated by GWHPs. This objective achievement will
Ebro River Basin (Spain). This basin constitutes the last stage in the
constitute an improvement in attaining a scientific-based management
evolution of the southern foreland basin of the Pyrenees (Barnolas and
tool, which allows the reproduction of real exploitation regimes and,
Robador, 1991; Pardo et al., 2004). The portion of the aquifer covering

156
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

the urban area of the city is known as ‘Urban Alluvial Aquifer of Zar- carried out by the owners by request of the local water management
agoza’ (Garrido et al., 2006, 2010a,b) due to the singular hydro- administrator (CHE, Spanish acronym). Later, datasets are transferred
geological properties of environment and groundwater use, among to the Geological Survey of Spain (IGME, Spanish acronym) for eva-
which geothermal exploitation is the most important. The Ebro River luation. The data logging consists of flow [l·s−1] and temperature [°C]
crosses the city in NW–SE direction, where two of its tributaries, the measurements at both extraction and injection wells, with a 15-min
Huerva and Gállego rivers, converge. The urban alluvial aquifer con- sampling frequency. The data registry shows a high variability between
sists of Early Pleistocene fluvial, alluvial, and aeolian sediments gen- the GWHP installations due to their different exploitation regimes. Data
erated by wide alluvial plains over an evaporite-dominated Miocene filtering and synchronization was performed based on a 15-min sam-
substratum (Quirantes, 1978; Simon et al., 2014). The surface of the pling frequency reference guide to obtain continuous and homogeneous
alluvial aquifer under the city is 632 km2 and its thickness ranges from data time series. In addition, temperatures registered while the opera-
5 to 60 m with a considerable spatial variability due to tectonic, grav- tion wells were inactive were discarded because the recorded values are
itational, diapiric, and dissolution processes affecting this detrital cover not representative.
(Sánchez-Navarro et al., 2004; Soriano et al., 2012). The transmissiv-
ities were derived from previous flow model calibrations (Garrido et al., 3.2. Application of different upscaling approximations
2006). This groundwater numerical model developed for the alluvial
aquifer of Zaragoza recreated the transient regime of groundwater flow The upscaling problem posed in the introduction consists in finding
considering a Cauchy boundary condition for the reproduction of the representative input variables at a given scale (e.g. daily scale) for
Ebro River hydraulic connection. This boundary condition is imposed groundwater flow and heat transport modelling obtained from an ex-
according to a time function providing the river head corresponding to haustive dataset at a smaller scale (e.g. 15-min scale) to reproduce
every time step. Hence, the flood events are considered in the model. adequately the operation of GWHP systems. It has to be stressed that
The western and eastern boundary conditions were chosen as pre- the term “scale” denoted on maps means exactly the opposite to the
scribed head or Dirichlet, corresponding to isohypses according to head scale definition used in upscaling and downscaling methods for en-
data measured in the field of the respective zones. The rest of the un- vironmental research.
assigned boundaries of the model were defined as Neumann boundary The operation of GWHPs is commonly implemented in a ground-
conditions with zero flux, which represent the lateral contact between water flow and heat transport model by assigning a prescribed water
the aquifer and the Tertiary bedrock. The results of the hydraulic test input/output rate and a prescribed temperature for each time step to
are in the range of 3·102–4·103 m2 day−1 (García-Gil et al., 2014b; the nodes of the model representing operation wells. Pumping-injection
Moreno, 2008). The depth of the phreatic surface ranges from 7 to 34 m rate (non-consumptive) is an additive, also called extensive, variable
and the main groundwater flow pattern is NW–SE in the southeastern that is proportional to the size of the sample. This feature makes the
part of the urban aquifer and W–E in the northwestern part or at the left upscaling of pumping-injection rates trivial. In contrast, defining a re-
margin of the Ebro River (García-Gil et al., 2015). presentative injection temperature becomes non-trivial. Temperature is
A total of 188 wells exploit the urban alluvial aquifer for geothermal an intensive variable, therefore it is not proportional to the sample size.
energy production (112 extraction wells and 76 injection wells). A To derive representative input injection temperatures, theoretical ex-
registry of 73 geothermic installations (65 in use) is currently gathered pressions or experience have to be considered. In this study, different
in an exploitation catalogue. The pumped groundwater in 2010 was upscaling formulas have been used, including daily quartiles (Q1, Q2,
estimated to be approximately 24·106 m3, with only 0.93·106 m3 being and Q3), geometric mean, and daily maximum and minimum values of
consumptive. The remaining withdrawal is associated with geothermal 15-minutal injection temperature datasets measured from injection
exploitation by means of GWHPs (Garrido et al., 2010b). The total heat wells. Also, a daily random registered injection temperature value was
power installed has been estimated to be 110 MWt of energy used for considered to evaluate the effect of a random approach on upscaling.
cooling purposes, with 21 installations being equipped with reverse- These upscaling approaches allowed covering the high variance of in-
cycle heat pumps with 34 MWt of the total heat power supplied jection temperature values, providing a basic-simple statistic approach
(Garrido et al., 2012a,b). The thermal groundwater regime of the city and laying the base framework for comparison of more complex ap-
and the extension of the heat plumes generated by GWHPs were re- proaches. On the other hand, in this work we propose the integration of
produced by numerical modelling of the groundwater flow and heat transferred energy by GWHP systems calculated at sampling frequency
transport (García-Gil et al., 2014a, 2015). scale to obtain a representative equivalent temperature (Teq) at a larger
A total of five GWHP geothermal exploitations were selected as scale adequate for efficient numerical modelling. The proposed calcu-
study objects in the present paper, corresponding to three hotels, a lation of this representative temperature is described by the following
museum, and a hospital in Zaragoza. The selection was carried out steps described hereafter. First, the energy transferred (Hm ) into the
based on different aspects, attempting to obtain the largest level of aquifer in the m time period of the sampling frequency is calculated by
representativeness. The oldest exploitation, which started in 1999, is the following expression:
SGS-4 and the most recent ones are SGS-1 and SGS-2, which started in
Hm = Qm Cw ρw (TIm−TCm) (1)
2010. All geothermal exploitations have two extraction wells and one
injection well, except for SGS-5, which has four extraction and three 3 −1
where Qm is the pumping rate [m ·s ] measured every m time
injection wells. The average depth below ground of these wells ranges period, c w is the specific heat capacity of water [J·kg−1·K−1], ρw is the
between 24 and 44 m. The average flow rate circulation of the in- water density [kg·m−3], (TI m−TCm) is the increment of the temperature
stallations SGS-1 to SGS-5 is 17.62, 11.4, 16.6, 19.16, and 50 l s−1, between the injection temperature (TI m ) and the extraction temperature
respectively. All studied installations have a climatization system that is (TCm ) during two measurements of m sampling period interval. In this
used both for cooling and heating, with an average heat power of 0.9, research, min, givingtheenergytransferredwiththeaquiferwitha min, giving
1.0, 1.9, 1.8, and 2.6 MW for SGS-1 to SGS-5, respectively. the energy transferred with the aquifer with a 15 min resolution. The
energy transferred (Hn ) into the aquifer in n period required in a nu-
3. Methodology merical model (time step) is given by:
n n
3.1. Origin and data treatment by filtering, synchronization, and Hn = ∑ Hm = ⎛⎜∑ Qm ⎞⎟ c w ρw (Teq + T¯Cn)
homogenization 1 ⎝ 1 ⎠ (2)
n n
The monitoring of the regime of geothermal system exploitations is where ∑ Qm = Qm1 + Qm2 + Qm3 + …+Qmn and ∑
1 1

157
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Hm = Hm1 + Hm2 + Hm3 + …+Hmn . In this paper, n = 96, which is the of a transient prescribed-flux and a transient prescribed-temperature
number of measurements to obtain daily (n = 1 day) integrated input boundary condition according to the time series measured and calcu-
values for the model. The equivalent temperature (Teq ) represents the lated by upscaling procedures. An initial temperature of 17 °C was
temperature value of injection in the model in order to transfer the chosen, approximately matching the annual mean of the region. A set of
energy transferred in to the aquifer throughout the n period (1 day in observation points at different distances from the injection wells (1, 3,
this case). The term T̄Cn stands for average captation temperature de- 5, 10, 20, 40, 100, and 200 m) was defined in all models used. The
fined as: observation points allowed to register the temperature evolution as a
n heat transport state variable for each upscaling optimization scenario
1
T¯Cn = ∑ TCm and to identify which calculated time series generated the closest
n 1 (3) thermal response to the real scenario (i.e., 15-min sampling frequency).
By rearranging Eq. (2), the following upscaling formula for The simulations were carried out under a transitory regime with a 15-
equivalent temperature (Teq) is obtained: min and one-day time step. For the sensitivity analysis of the proposed
upscaling procedure, a time step of 0.5, 2, and 5 days was adopted in
n
∑1 Hm two GWHP system models (SGS-1 and SGS-2).
Teq = n + T¯Cn
( ∑1 Qm) c w ρw (4)
3.5. Calculation of the upscaling procedure error
All the aforementioned upscaling approaches were used to calculate
daily time series for each of five GWHPs selected (Section 2) and were
The assessment of the deviation due to the proposed upscaling
subsequently implemented in a numerical groundwater flow and heat
procedures was carried out by comparing the reference scenario (15-
transport model for validation.
min measurements) with the calculated upscaling scenarios. The com-
parison was conducted using thermographs obtained from all defined
3.3. Sensitivity analysis of the upscaling process as a function of the time observation points. The temperature difference was obtained by sub-
resolution tracting the reference scenario from each optimization scenario to
calculate the deviation for all time steps. The deviation was calculated
To assess the upscaling procedure using the calculated equivalent as root-mean-square error (RMSE) using the following equation:
temperature, different temporal resolutions were considered (Δt ) . Three
n
time series considering half-daily, two-daily, and five-daily sampling ∑i = 1 d12
frequency were performed to evaluate the deviation in the evolution of RMSE =
n (5)
each scenario against 15-min measurement time series adopted as
realistic control time series. Furthermore, this analysis intended to ex- where d is the temperature difference between the optimization and
plore the possibilities of working with lower sampling frequencies to reference scenarios and n is the total number of existing values in each
correctly reproduce the exploitation regimes of the installations and considered set. The RMSE values approaching 0 were considered to
quantitatively check the different options available. reflect a high optimization degree of the upscaling procedure. During
the error analysis performed, special attention was focused on the
3.4. Numerical groundwater flow and heat transport models possible effects of the hydrogeological parameters on the RMSE dis-
tribution.
The FEFLOW code (Diersch, 2013) was used to reproduce 3D
groundwater flow and heat transport processes induced by five GWHP 4. Results
systems (SGS-1, SGS-2, SGS-3, SGS-4, and SGS-5; Fig. 2) in the alluvial
aquifer studied. The installations were selected based on different cri- 4.1. Results for the exploitation regime and upscaling procedures
teria, attempting to cover all existent ranges of hydraulic gradient,
permeability, and exploitation regimes. Therefore, they are considered The time series including the upscaling procedures proposed for
representative of the general operation of shallow geothermal systems each installation were calculated with the same approximation methods
located in the study area. A total of five local hydrogeological models but with different temporal resolutions. Fig. 4A shows the analysis
corresponding to the GWHP installations were implemented based on a carried out for installation SGS-1, where the 15-min sampling frequency
previous calibrated and validated 3D regional groundwater flow and series of the injection temperature was considered to be the real ex-
heat transport model covering the urban area of Zaragoza (Epting et al., ploitation regime. The registry period includes 273 days, with a mean
2017; Garcia-Gil et al., 2014). The modelled domain for each hydro- flow of 661 m3/day. The registry is continuously operating, except for
geological model analysed considered a > 600 m groundwater flux tube minor interruption intervals. The graphic shows that 55% of the time
to prevent border effects, with its geometry being obtained from a city- was utilized for cooling, thus causing a high-temperature injection
scale 3D regional model (Epting et al., 2017). A 40 m thick Miocene (> 25 °C), and the rest of the time was used for heating, leading to cold
basement and unsaturated zone (NSZ) were considered for downward water injection (< 15 °C). The calculated upscaling series of the injec-
and upward heat conduction, respectively. The domain characteristics tion temperature for the GWHP system SGS-1 (geometric mean, Q1, Q2,
of each installation model reflect different dimensions and mesh re- Q3, maximum, minimum, random value, and equivalent temperature)
finements depending on the number and spatial distribution of the for daily sampling frequency is shown in Fig. 4B–C.
GWHP operation wells. Each domain was discretized into an un- Regarding optimization, minimum data reduction of 98.9% was
structured finite element mesh with triangular elements (Table 1), all of achieved through upscaling transformation of daily scenarios (Table 3).
which were distributed in 12–14 layers depending on the model domain This generated a reduction in the computational time of 10.6 h on
(Fig. 3). A fixed-head boundary condition was imposed to the incoming average (Table 4).
and outlet groundwater flow nodes of the flux tube considered to re-
produce the groundwater flow, with hydraulic gradients of 1.35·10−4, 4.2. Results for the simulation of upscaling scenarios
8.60·10−4, 4.78·10−3, 7.27·10−4, and 1.04·10−2 for installations SGS-1
to SGS-5, respectively. A fixed ground temperature of 17 °C was as- The temperature distribution obtained from the reference time
signed to the left groundwater inlet boundary. The thermal and hy- series at the end of the simulations is shown in Fig. 5. The influence of
draulic parameterizations of the used models are listed in Table 2. The thermal dispersion and its interaction with established observation
boundary conditions defined for geothermal operation wells consisted points can be deduced based on the observation of heat plumes for the

158
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 2. The five studied GWHP exploitations (SGS-1, SGS-2, SGS-3, SGS-4, and SGS-5) and the distribution of the geothermal extraction and injection wells and
existing monitoring network. Projection: WGS 1984 Complex UTM Zone 30N.

Table 1
Characteristics of the domain modelled for each studied geothermal installa-
tion.
Installation Dimensions [m] Triangular Nodes Layers Period
elements [days]

SGS-1 710.4 × 658.2 × 30 29,052 16,341 12 273

SGS-2 847.7 × 592.9 × 30 859,392 467,116 12 241
SGS-3 772.6 × 647.8 × 30 88,928 48,315 14 264
SGS-4 672.7 × 513.9 × 30 45,192 24,885 14 310
SGS-5 663.3 × 469.2 × 30 56,994 31,200 14 212

last calculated time. The heat plume generated by installation SGS-1

(Fig. 5A) extends to a distance of 40 m, with a temperature increase of
1.75 °C with respect to the initial situation (18.75 °C isotherm). Fig. 5B
shows the thermal impact of installation SGS-2 on the aquifer, gen-
erating a larger heat plume than that of installation SGS-1 (23.79 °C Fig. 3. 3D finite element mesh used in the simulation of the geothermal ex-
isotherm, 40 m from the observation point), with a temperature incre- ploitation in a numerical groundwater flow and heat transport model. The ar-
ment of 3.84 °C at a distance of 100 m from the observation point rows represent the groundwater flux tube limited by the basement and un-
(20.84 °C isotherm). Installation SGS-3 generated the largest plume saturated zone (NSZ).
(Fig. 5C), with the groundwater temperature at the 40 m observation
point increasing to an isotherm of 24.31 °C and at the furthest ob-
servation point of 200 m increasing by 3.65 °C (20.65 °C isotherm). The

159
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Table 2 Table 3
Thermal and hydraulic parameterization of the domains for the five analysed Efficiency of the upscaling procedure. Total number of measurements registered
installations. for each geothermal system studied with 15-min sampling interval and final
number of data obtained from the calculated upscaling procedures.
Parameter Basement Aquifer NSZ
GWHP Number of measurements Number of data Efficiency of
Hydraulic conductivity [m·day−1] 10−7 188–963 10−7 Installation with 15-min sampling obtained data reduction
Specific storage coefficient [m−1] 10−3 5·10−3–0.3 0,30 interval acquired through [%]
Dynamic porosity [–] 0,01 0,30 0,30 upscaling
Volumetric heat capacity [MJ·m−3·K−1] 1,07 2,92 1,73
Thermal conductivity [W·m−1·K−1] 1,26 2,10 1,80 SGS-1 26,302 273 98.9
Longitudinal dispersivity [m] 5,00 4,93 5,00 SGS-2 23,178 241
Transversal dispersivity [m] 1,00 1,95 1,00 SGS-3 25,439 264
SGS-4 29,855 310
SGS-5 20,442 212
thermal impact caused by installation SGS-4 is shown in Fig. 5D. In-
stallation SGS-4 generated a heat plume with a temperature increment
of 3.55 °C at a distance of 40 m (20.55 °C isotherm). The temperature Table 4
increased by only 0.82 °C (17.82 °C isotherm) at a distance of 100 m. Simulation time of the real data series of the injection temperature and the
upscaling approaches with daily frequency.
The temperature effect is almost negligible at this observation point.
The smallest heat plume was generated by installation SGS-5 (Fig. 5E), Installation Simulation of series 15-min Simulation of daily series
with an observed temperature increment of only 0.43 °C (17.43 °C iso- [hours] [minutes]
therm) at a 40 m distance from the injection point.
SGS-1 10 3
In addition, the thermographs obtained from reference scenario si- SGS-2 16 6
mulations are shown in Fig. 6 for all GWHP systems. The temperature SGS-3 11 6
evolution of the reference scenarios and each upscaling procedure for SGS-4 8 3
SGS-5 8 3
all time steps simulated at 20 m from the injection point for each GWHP
installation are shown in Fig. 7 (all other distances are shown for SGS-3
as supplementary material, Fig. S1). The temperature evolution for each
largest upscaling error occurs when using the minimum-value upscaling
of the upscaling procedures and their difference with respect to the
approach, resulting in a 1 °C to 5.5 °C error. The deviation of the
measured reference values can be observed in these thermographs. The
maximum value approach ranges from 1.2 °C to 3.5 °C and that of Q1
upscaling approaches that cause the largest error are associated with
and Q3 approaches results in an error of approximately 2 °C. The Q2,
scenarios in which maximum and minimum values were incorporated,
geometric mean, and random value approaches show a similar devia-
deviating up to 5 °C from the real measured temperature scenario in
tion of 1.5 °C. The equivalent calculated temperature is the only up-
some exploitations (average deviation of ± 3 °C). The same phenom-
scaling procedure with a deviation below 0.8 °C for all GWHP in-
enon occurs when using Q1 and Q3: although the deviation is smaller
stallations. In addition, the average RMSE values of all installations for
(average deviation of ± 1.8 °C), it does not follow the real regime of the
all upscaling procedures are presented in Fig. 8. The results show that
studied installations. Regarding the data series of the geometric mean,
the minimum-value upscaling approach produces the largest error,
Q2, and random value, the deviation decreases to an average value
ranging from 0.8 °C to 4.6 °C. In contrast, scenarios using the equivalent
of ± 1 °C.
calculated temperature approach present the lowest error range:
The equivalent calculated temperature scenario causes the least
0.3–0.5 °C.
deviation ( ± 0.4 °C) with respect to the reference scenarios for all
studied installations.
4.4. Error calculation for the upscaling procedures as a function of the time
resolution
4.3. Upscaling procedure error calculation
The error (RMSE) associated with the use of the equivalent tem-
The distribution of upscaling errors (RMSE) for each of the GWHP perature series carried out in all studied installations with daily tem-
installations at all distances evaluated can be observed in Fig. 8. The poral frequency was calculated along different observation points

Fig. 4. Thermograph of installation SGS-1 showing the real injection temperature obtained form 15-min sampling frequency (A). Minimum and maximum (B); Q1,
Q2 and Q3 (C); Random and geometric mean (D) and equivalent temperature (E) Upscaled daily injection temperatures calculated.

160
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 5. Isoline map of each studied installation with thermic plumes generated for the last calculated time. Status of the extraction and injection wells and observation
points established in the model. A: SGS-1. B: SGS-2. C: SGS-3. D: SGS-4. E: SGS-5.

(Fig. 9). installations and each of the upscaling scenarios decreases to approxi-
In addition, the RMSE values derived from the equivalent tem- mately 1 °C. For half-daily and daily resolution scenarios, a maximum
perature upscaling approach for all observation points established in deviation of 0.25 °C was obtained at the first (10 m) observation point.
the numerical models (SGS-1 and SGS-2) with different temporal fre-
quency (half-daily, two-daily, and five-daily) are shown in Fig. 10. A 5. Discussion
high deviation was observed for the five-daily sampling frequency
scenario for both installations, with a deviation of 2.1 °C and 2.5 °C, The equivalent temperature upscaling approach calculated by the
respectively, at the nearest observation point (1 m) from the injection integration of the daily dissipated heat power by a GWHP installation is
well. The two-daily sampling frequency scenario caused a deviation considered to provide the highest reliability because it is derived from
similar to that of the daily sampling frequency scenario, with RMSE the real operation regime of the installation. Using the equivalent
values of 1.6 °C and 1.9 °C, respectively. The daily and half-daily sam- temperature approach scenarios, the time series (half-daily, two-daily,
pling frequency scenarios presented a similar average deviation of and five-daily sampling frequency) were obtained for two of the studied
1.5 °C. On the other hand, if observation points located near the in- installations, where the simulation time was almost the same as that of
jection well (1, 3, and 5 m) are ignored, the deviation for both the previous case (daily sampling frequency approach).

161
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 6. Thermographs based on the simulation of the injection temperature series with 15-min sampling frequency distributed along the observation point for each of
the studied installations. This is the reference for the real exploitation regime in each analysis.

Because the exploitation regime of geothermal installations does not injection temperatures compared to those of the other exploitations
follow a common pattern, each one was analysed individually. In (< 30 °C).
general, SGS-1, SGS-3 and SGS-5 had a continuous operation regime Based on the numerical model simulations of the different scenarios,
with the highest average flows (661, 865, and 892 m3/day, respec- the error derived from the upscaling approximations of daily sampling
tively), except for time intervals during which they did not inject water. frequency against high real sampling frequency data was quantified.
On the contrary, installations SGS-2 and SGS-4 developed a variable The temperatures generated along observation points allowed the de-
regime with lower flows (457 and 265 m3/day) and prolonged intervals duction of the deviation evolution. Daily approximations calculated
in which they were off. The climatization demand is only for cooling based on the minimum and maximum caused the widest average de-
purposes in SGS-3 and SGS-4, a fact that differentiates them from the viation range, varying from 0.8 °C to 4.6 °C and from 1.2 °C to 3.2 °C,
rest of the installations, which have a combined use for both heating respectively. The rest of the statistic parameters had a global average
and cooling purposes. The plumes that reached the largest thermic in- standard deviation of 0.73 °C. The smallest deviation corresponded to
fluence area (100 and 200 m) were generated by installations SGS-2 and the equivalent temperature scenario and ranged between 0.3 °C and
SGS-3, recording the highest injection temperatures (> 30 °C) during 0.5 °C. This means that the calculated temperature is the best and most
summertime. Installation SGS-4 generated a smaller plume due to the useful upscaling approximation because it is the only parameter that
injection of water at a lower temperature (< 30 °C) and its variable generated a deviation range below 0.8 °C among the analysed in-
regime, remaining off during large periods of time. The smallest mea- stallations.
sured plumes were generated by installations SGS-1 and SGS-5, derived On the other hand, the approximations carried out with different
from a combination of cooling and heating. Also, they presented lower temporal resolutions using the equivalent temperature upscaling

162
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 7. Thermographs representing the real series of the injection temperature (15-min) before calculating the upscaling with the daily sampling frequency for each
studied exploitation. The 20 m observation point is the only point considered to obtain the best visualization and a representative distance for the evaluation of
possible impacts.

approach in two of the analysed installations proved that both half- intervals is not particularly useful. Furthermore, ignoring observation
daily and daily scenarios considering all observation points generated points near the injection wells (1, 3, and 5 m) led to a 1 °C reduction in
the lowest deviation values. Thus, reducing the established temporal the error range. The maximum deviation was observed at the

Fig. 8. Distribution of upscaling approach errors (RMSE) for all analysed installations (SGS-1, SGS-2, SGS-3, SGS-4, and SGS-5) and the average RMSE of each
upscaling approach represented by error bars.

163
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Fig. 9. Graphical representation of the RMSE of the equivalent temperature data series with daily frequency for all studied installations considering that all ob-
servation points are situated at different distances.

regimes in regional models at the city scale. Furthermore, the validation

of the proposed methodology based on standard hydrogeological
parameters ensures its transferability to other equivalent hydro-
geological environments, that is, shallow alluvial aquifers, which are
prevailing in urban environments. The error analysis performed on the
upscaling procedures has not highlighted any relationship between the
hydrogeological parameters and the errors observed. This indicates that
the upscaling procedure proposed is significantly independent from the
hydrogeological parameters in the parameter ranges evaluated in this
study. Furthermore, when applying the results found to other hydro-
geologically different city aquifers, it has to be noted that the optimal
upscaling approach might be influenced by the relevance of advection
versus conduction heat transport mechanisms given by the Peclet
number (Domenico and Schwartz, 1998), which relates the energy
transported by fluid motion to the energy transported by conduction. In
the cases evaluated, heat advection is greater than heat conduction,
which is a situation likely found in alluvial aquifers. Situations where
transmissivities and hydraulic gradients are lower than those con-
sidered in this work are not suitable for the application of the conclu-
sions found in this research.
Conversely, although the equivalent temperature upscaling proce-
dure has been shown as the best approach to reproduce the thermal
pollution signal of a GWHP, other approaches might be of interest for
authorities. When considering the maximum impact produced by a
Fig. 10. Calculation of the RMSE as a function of the temporal resolution of GWHP in terms of regulatory requirements or derogation of other users
equivalent temperature data series in two of the analysed installations (A: SGS- from the resource, regulators might consider using daily maximum
1 and B: SGS2). temperature as a conservative approach to integrate the high frequency
sampling data. This approach would tend to overestimate thermal
plumes generated but would be safer in case of future peak conflicts.
observation point at 10 m distance (0.25 °C). This is due to the fact that
GWHP systems are likely to present similar temporal demand profiles
the points near the GWHP operation wells are affected by hourly
but their interpolation to longer times might get to losing peak de-
thermal exploitation regimes, thus indicating that a daily or even higher
mands. This perspective indicates that the equivalent temperature
sampling frequency simulation is necessary to reproduce a thermal
proposed does not have to be necessarily used in numerical models for
regime with a reasonable error at points close to the injection well.
the assessment of thermal impacts required for thermal discharge au-
Considering that the evaluation of thermal impacts close to the injec-
thorizations.
tion wells of GWHP systems is not the aim of regional-scale models and
that the mean minimum distance of GWHP wells to the monitoring
points for Zaragoza is, for all cases, greater than 20 m, the errors de- 6. Conclusions
rived from the upscaling procedure proposed at the first 20 m from the
injection well could be neglected. Based on the research carried out in The exploitation regime of geothermal systems presents a highly
this paper, the obtained results are of interest to obtain time functions variable pattern that increases the difficulty of processing data from
of daily sampling frequency, even when their operation regime is highly such installations prior to their implementation in regional-scale
variable, thus enabling the implementation of these exploitation groundwater flow and heat transport models. Monitoring of shallow
geothermal systems is necessary and should be mandatory to guarantee

164
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

the long-term exploitation of the renewable energy resources and the References
aquifer sustainability. The research performed in this paper revealed
the difficulties associated with the necessary treatment to study the Banks, D., 2009. Thermogeological assessment of open-loop well-doublet schemes: a re-
exploitation regime of GWHP installations. Based on daily approxima- view and synthesis of analytical approaches. Hydrogeol. J. 17 (5), 1149–1155.
Barnolas, A., Robador, A., 1991. Geological Map of Spain N° 383 Zaragoza. Geological
tions, calculated by upscaling of the injection temperature, a data re- Survey of Spain (IGME), Madrid (Spain).
duction of 98.9% was obtained. This makes an adequate implementa- Bayer, P., Saner, D., Bolay, S., Rybach, L., Blum, P., 2012. Greenhouse gas emission
tion of GWHP systems in hydrogeological models possible, thus making savings of ground source heat pump systems in Europe: a review. Renew. Sustain.
Energy Rev. 16 (2), 1256–1267.
them more efficient because a large amount of simulation time is saved, Bedrikovetsky, P., 2008. Upscaling of stochastic micro model for suspension transport in
accounting for a decrease from an average duration of 10 h to 4 min. porous media. Transp. Porous Media 75 (3), 335–369.
The upscaling scenarios calculated by descriptive statistic parameters Bierkens, M.F.P., van der Gaast, J.W.J., 1998. Upscaling hydraulic conductivity: theory
and examples from geohydrological studies. In: Finke, P.A., Bouma, J., Hoosbeek, M.
(geometric mean, Q1, Q2, Q3, maximum, and minimum), together with R. (Eds.), Soil and Water Quality at Different Scales: Proceedings of the Workshop
a random approach and the equivalent injection temperature proposed “Soil and Water Quality at Different Scales” held 7–9 August 1996, Wageningen, The
in this paper allowed to obtain a global approximation to the variability Netherlands. Springer Netherlands, Dordrecht, pp. 193-207.
Bierkens, M., Finke, P., De Willigen, P., 2000. Upscaling and Downscaling Methods for
range of the injection temperatures and to calculate the deviation for
Environmental Research. Kluwer Academic.
each scenario. By making use of the proposed novel upscaling proce- Brus, D.J., de Gruijter, J.J., 1997. Random sampling or geostatistical modelling? Choosing
dure proposed here to calculate a daily equivalent temperature from the between design-based and model-based sampling strategies for soil (with discussion).
heat power transferred by real groundwater heat pump installations, Geoderma 80 (1), 1–44.
Dagan, G., 1981. Analysis of flow through heterogeneous random aquifers by the method
with both daily and half-daily sampling frequencies, it is possible to of embedding matrix: 1. Steady flow. Water Resour. Res. 17 (1), 107–121.
reduce significantly the amount of data required to optimize its im- de Vries, W., Kros, J., van der Salm, C., Groenenberg, J.E., Reinds, G.J., 1998. The use of
plementation in complex numerical models at an urban scale. The upscaling procedures in the application of soil acidification models at different spatial
scales. Nutr. Cycl. Agroecosyst. 50 (1), 223–236.
methodology has been validated on five different real scenarios cov- Diersch, H.J., 2013. FEFLOW: Finite Element Modeling of Flow, Mass and Heat Transport
ering the most standard hydrogeological conditions in shallow alluvial in Porous and Fractured Media. Springer, Berlin Heidelberg.
aquifers to ensure its transferability. Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology. Wiley.
Epting, J., Händel, F., Huggenberger, P., 2013. Thermal management of an un-
The most relevant conclusions of this study with respect to nu- consolidated shallow urban groundwater body. Hydrol. Earth Syst. Sci. 17 (5),
merical models of the heat regime in urban aquifers with shallow 1851–1869.
geothermal installations are the following: Epting, J., García-Gil, A., Huggenberger, P., Vázquez-Suñe, E., Mueller, M.H., 2017.
Development of concepts for the management of thermal resources in urban areas –

• Recording temperature and pumping rates with a 15-min sampling

assessment of transferability from the Basel (Switzerland) and Zaragoza (Spain) case
studies. J. Hydrol. 548, 697–715.
frequency or higher is convenient to estimate the thermal power Epting, J., Huggenberger, P., 2013. Unraveling the heat island effect observed in urban
groundwater bodies–definition of a potential natural state. J. Hydrol. 501, 193–204.
dissipated by shallow geothermal exploitations due to the poten-
Finke, P.A., Bierkens, M.F., de Willigen, P., 2002. Choosing Appropriate Upscaling and
tially high variation of the exploitation regime over time. Downscaling Methods for Environmental Research. International Association of
• The upscaling procedure that best fits real operations of geothermal Hydrological Sciences Publication, pp. 405–409 (vol. 273).
Galgaro, A., Cultrera, M., 2013. Thermal short circuit on groundwater heat pump. Appl.
systems is the calculation of the equivalent injection temperature by
Therm. Eng. 57 (1–2), 107–115.
integrating the daily dissipated energy. Garcia-Gil, A., Vazquez-Sune, E., Schneider, E.G., Sanchez-Navarro, J.A., Mateo-Lazaro,
• Regional-scale numerical modelling of all studied exploitation re- J., 2014. The thermal consequences of river-level variations in an urban groundwater
body highly affected by groundwater heat pumps. The Science of the total environ-
gimes can be efficiently implemented throughout daily resolution
ment 485–486, 575–587.
time-series data. García-Gil, A., Vázquez-Suñe, E., Garrido, E., Sánchez-Navarro, J.A., Mateo-Lázaro, J.,
• The error analysis indicates that the error values of the calculated 2014a. The thermal consequences of river-level variations in an urban groundwater
body highly affected by groundwater heat pumps. Sci. Total Environ.
thermographs are lower than 0.4 °C ± 0.1 °C using the proposed
García-Gil, A., Vázquez-Suñe, E., Schneider, E.G., Sánchez-Navarro, J.Á., Mateo-Lázaro,
upscaling procedure presented in this work. These results add ad- J., 2014b. The thermal consequences of river-level variations in an urban ground-
ditional uncertainty to the calibration process of numerical models water body highly affected by groundwater heat pumps. Sci. Total Environ. 485,
and therefore should be considered when estimating thermal para- 575–587.
García-Gil, A., Vázquez-Suñe, E., Schneider, E.G., Sánchez-Navarro, J.Á., Mateo-Lázaro,
meters.
J., 2015. Relaxation factor for geothermal use development – criteria for a more fair
and sustainable geothermal use of shallow energy resources. Geothermics 56,
In conclusion, the present paper depicts an advance in the im- 128–137.
Garrido, E., Arce, M., Van Ellen, W., 2006. Modelo matemático de flujo subterráneo del
plementation process of real shallow geothermal systems in numerical
acuífero aluvial del Ebro en el entorno de Zaragoza. De la toma de datos y la
heat flow models and provides an enhancement of the optimal man- realización de modelos de agua subterránea a la gestión integrada. serie
agement of renewable energy resources in urban environments. Hidrogeología y Aguas Subterránea, vol. 21, pp. 343–349.
Garrido, E., Sánchez Navarro, J.A., Coloma, P., 2010a. Geothermic Use of the Urban
Alluvial Aquifer of Zaragoza: First Results.
Acknowledgments Garrido, E., García-Delanoceda, C., Moreno, L., Sánchez-Navarro, J., 2012a. Criterios
técnicos e hidrogeológicos para la explotación del acuífero urbano de Zaragoza por
The authors would like to thank the anonymous reviewers for their sistemas geotérmicos abiertos. In: Madrid, F.d.l.E.d.l.C.d. (Ed.), III Congreso de
Energía Geotérmica en la Edificación y la Industria. Fundación de la Energía de la
valuable comments and suggestions to improve the quality of the paper. Comunidad de Madrid, Madrid, pp. 495–504.
This work was performed by the Geological Survey of Spain (IGME) Garrido, E., Oroz, M., Coloma, P., Sánchez-Navarro, J., 2012b. Transformación de usos y
under the Cooperation Agreement Framework between IGME and the demandas del acuífero urbano de Zaragoza derivado del aprovechamiento
geotérmico de sus aguas subterráneas. In: AIH-GEAIH-GE (Ed.), Las aguas
Ebro Hydrographic Confederation (CHE) for the specific arrangement subterráneas: desafíos de la gestión para el siglo XXI. International Association of
entitled ‘The application of a groundwater flow and heat transport Hydrogeologists, Zaragoza, pp. 7.
numerical model for the simulation of management strategies of geo- Garrido, E., García-Gil, A., Vázquez-Suñè, E., Sánchez-Navarro, J.Á., 2016. Geochemical
impacts of groundwater heat pump systems in an urban alluvial aquifer with eva-
thermal installations in the city of Zaragoza’. Alejandro García-Gil
poritic bedrock. Sci. Total Environ. 544, 354–368.
gratefully acknowledges MIKE by DHI for the sponsored FEFLOW li- Garrido, E.A., Sánchez-Navarro, J.A., Coloma, P., 2010b. Geothermic use of the urban
cense. alluvial aquifer of Zaragoza: first results. Geogaceta Geol. Soc. Spain 49, 4.
Gropius, M., 2010. Numerical groundwater flow and heat transport modelling of open-
loop ground source heat systems in the London Chalk. Q. J. Eng. Geol. Hydrogeol. 43
Appendix A. Supplementary data (1), 23–32.
Herbert, A., Arthur, S., Chillingworth, G., 2013. Thermal modelling of large scale ex-
Supplementary data associated with this article can be found, in the ploitation of ground source energy in urban aquifers as a resource management tool.
Appl. Energy 109, 94–103.
online version, at https://doi.org/10.1016/j.jhydrol.2018.05.057.

165
S. Muela Maya et al. Journal of Hydrology 563 (2018) 155–166

Jaudin, F., 2013. D2.2: General Report of the current situation of the regulative frame- Simon, J.L., et al., 2014. Interacting tectonic faulting, karst subsidence, diapirism and
work for the SGE systems. Regeocities, pp. 50. continental sedimentation in Pleistocene deposits of the central Ebro Basin (Spain).
Johannes Dolman, A., Blyth, E.M., 1997. Patch scale aggregation of heterogeneous land Geol. Mag. 151 (6), 1115–1134.
surface cover for mesoscale meteorological models. J. Hydrol. 190 (3), 252–268. Soriano, M.A., et al., 2012. Quaternary alluvial sinkholes: record of environmental con-
Lo Russo, S., Gnavi, L., Roccia, E., Taddia, G., Verda, V., 2014. Groundwater Heat Pump ditions of karst development. Examples from the Ebro Basin, Spain. J. Cave Karst
(GWHP) system modeling and Thermal Affected Zone (TAZ) prediction reliability: Stud. 72 (2), 173–185.
influence of temporal variations in flow discharge and injection temperature. Spitler, J., 2005. Editorial: Ground-source Heat Pump System Research—Past, Present,
Geothermics 51, 103–112. and Future. Taylor & Francis.
Lund, J.W., Boyd, T.L., 2016. Direct utilization of geothermal energy 2015 worldwide Viscarra Rossel, R.A., Brus, D.J., Lobsey, C., Shi, Z., McLachlan, G., 2016. Baseline esti-
review. Geothermics 60, 66–93. mates of soil organic carbon by proximal sensing: comparing design-based, model-
Moreno, L.M., 2008. Hidrogeología Urbana de Zaragoza. IGME. assisted and model-based inference. Geoderma 265, 152–163.
Pardo, G., 2004. La Cuenca del Ebro. In: Vera, J.A. (Ed.), Geología de España. Sociedad Vogler, D., Ostvar, S., Paustian, R., Wood, B.D., 2018. A hierarchy of models for simu-
Geológica de España-Instituto Geológico y Minero de España, pp. 533–543. lating experimental results from a 3D heterogeneous porous medium. Adv. Water
Parsons, M.L., 1970. Groundwater thermal regime in a glacial complex. Water Resour. Resour. 114, 149–163.
Res. 6 (6), 1701–1720. Whitaker, S., 1986. Flow in porous media I: a theoretical derivation of Darcy's law.
Quirantes, J., 1978. Estudio sedimentolôgico y estratigâfico del terciario continental de Transp. Porous Media 1 (1), 3–25.
los monegros (Sedimentological and stratigraphical study of the continental Tertiary Wu, J., Jones, K.B., Li, H., Loucks, O.L., 2006. Scaling and Uncertainty Analysis in
of the Monegros region). Diputaciôn Provincial de Zaragoza. Ecology: Methods and Applications. Springer, Netherlands.
Sánchez-Navarro, J., Jiménez, N., Galve, J., Asta, M., 2004. Estudio hidrogeológico de la Yu, Y., Bathurst, R.J., Damians, I.P., 2016. Modified unit cell approach for modelling
subsidencia y los humedales salinos en el entorno de Zaragoza. Unidad de Montes y geosynthetic-reinforced column-supported embankments. Geotext. Geomembr. 44
Áreas Naturales, Excmo. Zaragoza, Ayuntamiento de Zaragoza Unpublished Report. (3), 332–343.

166