Professional Documents
Culture Documents
Research article
a r t i c l e i n f o a b s t r a c t
Article history: The assessment of the efciency of wastewater treatment plants (WWTPs) is essential to compare their
Received 5 October 2015 performance and consequently to identify the best operational practices that can contribute to the
Received in revised form reduction of operational costs. Previous studies have evaluated the efciency of WWTPs using con-
16 November 2015
ventional data envelopment analysis (DEA) models. Most of these studies have considered the opera-
Accepted 19 November 2015
tional costs of the WWTPs as inputs, while the pollutants removed from wastewater are treated as
outputs. However, they have ignored the fact that each pollutant removed by a WWTP involves a
different environmental impact. To overcome this limitation, this paper evaluates for the rst time the
Keywords:
Economic savings
efciency of a sample of WWTPs by applying the weighted slacks-based measure model. It is a non-radial
Efciency DEA model which allows assigning weights to the inputs and outputs according their importance. Thus,
Performance the assessment carried out integrates environmental issues with the traditional techno-economic ef-
Shadow price ciency assessment of WWTPs. Moreover, the potential economic savings for each cost item have been
Weighted slacks based measure quantied at a plant level. It is illustrated that the WWTPs analyzed have signicant room to save staff
and energy costs. Several managerial implications to help WWTPs' operators make informed decisions
were drawn from the methodology and empirical application carried out.
2015 Elsevier Ltd. All rights reserved.
1. Introduction Within the urban water cycle, special attention has been paid to
the efciency assessment of wastewater treatment plants
The adoption of regulations such as European Union (EU) (WWTPs). In particular, assessing the efciency of WWTPs allows
Directive 91/271/ECC concerning urban wastewater or the United their performance to be compared and therefore best practices to
States Clean Water Act has involved signicant progress in the be identied. In this context, Hern
andez-Sancho et al. (2011), Sala-
development and implementation of wastewater treatment tech- Garrido et al. (2012a), Molinos-Senante et al. (2015a), and Guerrini
nologies. However, water resource issues should not be addressed et al. (2015), among others, have assessed the so-called techno-
only from a technical and engineering perspective, but they should economic efciency of WWTPs. In doing so, they have considered
have a multidisciplinary nature, as the EU Water Framework that the operational and maintenance costs of WWTPs are their
Directive (WFD) published in 2000 highlighted. Accordingly, the inputs, while the pollutants removed from the wastewater are their
performance assessment of water facilities should involve eco- outputs. This assessment is very useful, as it allows economic and
nomic, environmental, and technical issues. technical variables to be integrated into a single indicator, namely
the efciency index (Molinos-Senante et al., 2014). However, they
ignored the fact that the different pollutants removed from
* Corresponding author. bDepartamento de Ingeniera Hidr aulica y Ambiental, wastewater would involve different impacts on the environment if
lica de Chile, Av. Vicun
Ponticia Universidad Cato ~ a Mackenna 4860, Santiago, Chile. they were dumped.
E-mail addresses: ll.castellet@gmail.com (L. Castellet), mmolinos@uc.cl From a methodological point of view, almost all previous studies
(M. Molinos-Senante). have applied the data envelopment analysis (DEA) approach to
1
Both authors contributed equally to this work.
http://dx.doi.org/10.1016/j.jenvman.2015.11.037
0301-4797/ 2015 Elsevier Ltd. All rights reserved.
L. Castellet, M. Molinos-Senante / Journal of Environmental Management 167 (2016) 160e166 161
evaluate the efciency of WWTPs. As Zhu (2015) stated, in DEA pollutants removed from wastewater as outputs, none of these
models the efcient frontier is obtained empirically from a real previous studies have weighted the outputs based on their envi-
database of the consumed resources (inputs) and generated prod- ronmental impacts. Moreover, no studies were identied that
ucts (outputs). Those units which are located over the frontier mark quantify the potential economic savings for each cost item that
the minimum number of inputs that a WWTP could operate being WWTPs could achieve if they were efcient. In this sense, Molinos-
efcient (input orientation) or the maximum production achieved Senante et al. (2014) computed an efciency score for each input
given a certain number of resources (output orientation) (Azad and (cost item), but they did not quantify the potential economic sav-
Ancev, 2014). Broadly, two types of DEA models can be distin- ings associated with each one.
guished, namely radial and non-radial models. They differ, among From a policy perspective, the results of this study are of great
other things, in how the distance between the units is analyzed and interest for WWTP managers. On the one hand, the evaluation of
the production frontier is measured. In radial models, the distance the efciency of WWTPs allows managers to compare the perfor-
from the units to the production frontier is obtained by projecting mance of the set of facilities assessed. Hence, best operational
on the efcient frontier the line that joins the analyzed decision- practices could be identied. The implementation of these practices
making unit (DMU) with the origin (Carvalho and Marques, in inefcient WWTPs would contribute to the improvement of their
2011). Each output and input that constitutes the virtual input performance and consequently the reduction of their operational
and output, respectively, contributes differently in obtaining the costs. On the other hand, the quantication of potential economic
virtual input and output. We refer to the input and output contri- savings for each cost item is essential to support the decision-
butions as weights. The weights corresponding to each input and making process. Thus, WWTP managers could implement the
output are given by the program itself so that the efciency of the most appropriate measures to further reduce operational costs.
unit is maximized (Cooper et al., 2011).
A signicant shortcoming of radial DEA models is that they as-
sume that the reduction or increase of inputs and outputs are
proportional, which is not always the case in real examples. To
overcome the problem mentioned before, Tone (2001) proposed 2. Methodology
the slacks-based model (SBM) which is a non-radial DEA model
that aims to individually optimize each resource and/or product 2.1. Efciency assessment
that consumes or produces each production unit. This efciency
score is obtained for each input/output, so that we can see in detail To compute the efciency of WWTPs, the WSBM model intro-
the behavior of each input/output. duced by Tone (2011) was applied as it has the advantages of non-
In most of the empirical applications, the fact that DEA meth- radial DEA models and also it allows us to assign weights to the
odology automatically assigns the weights to each input and output inputs and outputs for all the DMUs according to their importance
is an advantage, as it reduces the subjectivity of the assessment or relevance from an environmental point of view. In other words,
(Murillo-Zamorano, 2004). However, depending on the priorities or the WSBM model allows the integration of environmental issues in
managerial preferences given to the outputs or even depending on the assessment of the efciency of WWTPs.
the own outputs of the DMUs, it is more suitable to assign specic Let the sample of WWTPs to be evaluated as J{1, 2, , n}, each
weights to the outputs (Barros et al., 2012). In the case of WWTPs, it WWTP having m inputs and s outputs. The vector of inputs and
is well known that the negative impact that different pollutants outputs for the WWTPj is denoted by xj(x1j, x2j, , xmj) and yj(y1j,
would involve in the environment if they were not removed from y2j, , ysj), respectively. The production possibility set is dened as
wastewater is also different. Hence, to integrate the environmental follows (Hoang, 2014):
impact in the efciency assessment of WWTPs, it is necessary to
8 9
manually assign the weight to each output rather than them < =
X
n X
n
being assigned through optimization. This can be done by P x; yjxi xij ci; 0 yr yrj cr; el 1; l 0
computing the efciency using the weighted slack-based measure : ;
j1 j1
(WSBM) model introduced by Tone (2011). The model is a non-
(1)
radial DEA model similar to the SBM model, but it allows weights
to be assigned manually to the set of inputs and outputs.
where e denotes a row vector in which all elements are equal to one
Against this background, the objectives of this paper are
and l(l1, l2, , ln) is an intensity vector.
twofold. The rst is to assess the efciency of a sample of WWTPs
The introduction of the slacks allows the inequalities of Eq. (1) to
by integrating technical, economic, and environmental issues. In
be transformed into the following equalities (Tone, 2010):
doing so, the non-radial WSBM model is applied, since it allows us
to assign weights to the set of outputs produced by the WWTPs
X
n
(pollutants removed from wastewater) according their environ- x lj xj s (2)
mental impact. Moreover, since energy use in WWTPs involves not j1
only economic issues but also environmental aspects, this input
was also weighted based on the shadow price of CO2 emissions
from WWTPs. Moreover, and as a sensitivity analysis, the impact of X
n
y lj yj s (3)
different energy weights on the efciency of the WWTPs was
j1
evaluated. The second objective is to quantify the potential eco-
nomic savings for each cost item at the WWTP level.
This paper contributes to the current strand of literature in the s 0; s 0 (4)
eld of WWTP performance measurement by computing the ef-
ciency scores of a sample of WWTPs by introducing weights to the where s s m s
1 ; s2 ; ; sm 2R and s s1 ; s2 ; ; ss 2R are
pollutants removed from wastewater which represent its impact on the input and output slacks, respectively.
the environment. It should be highlighted that in spite of some According to Tone (2011), the WSBM model is dened as
studies evaluating the efciency of WWTPs considering the follows:
162 L. Castellet, M. Molinos-Senante / Journal of Environmental Management 167 (2016) 160e166
Regarding the selection of the outputs, two approaches have WWTPs do not need to improve their performance, as they are
previously been considered, namely the volume of treated water identied as the units with the best practices. Improvement efforts
(Gjebrea and Zoto, 2013; Lannier and Porcher, 2014) and the should be focused on inefcient WWTP units, which represent 59%
quantity of contaminants removed from the wastewater. As Wu of the sample; the results also change dramatically, giving an
et al. (2015) pointed out, the concentration of pollutants in the average efciency index of 0.2784 for this group of WWTPs. This
inuent depends on several environmental and climatic factors, nding suggests that the saving potential is much greater when
while the efuent of all WWTPs must meet the legal requirements. only inefcient WWTPs are contemplated, meaning that these
Hence, the concentration of pollutants to be removed from inefcient WWTPs could save as much as 72% of their costs while
wastewater is not the same for all WWTPs. Because the operational maintaining the quality of the treated water.
costs of WWTPs are affected by the efciency of the pollutant Fig. 1 shows the plants (score equal to 1) that comprise the
removal (Rodriguez-Garcia et al., 2011), it was considered that four efcient frontier (or benchmark) of best practices in relation to the
pollutants constitute the outputs obtained from the treatment efcient WWTPs, and Fig. 2 groups the WWTPs in terms of ef-
process: (i) suspended solids (SS); (ii) organic matter measured as ciency scores. It is illustrated that 20 out of the 49 WWTPs evalu-
chemical oxygen demand (COD); (iii) nitrogen (N); and (iv) phos- ated are efcient. On the other hand, the remaining plants (29 out
phorus (P). of 49) have an efciency score lower than 0.6. This nding evi-
Table 1 provides a snapshot of the statistical data used to dences the great potential for performance improvement in inef-
compute the efciency scores for each WWTP. It is illustrated that, cient WWTPs. Fig. 2 illustrates that there are no plants with
on average, the most important cost item is staff, representing 36% moderate inefciency, as none of the 49 plants assessed has an
of total costs. Energy is the next in importance, representing 22% of efciency score ranged between 0.60 and 0.99. From a manage-
total costs. Waste management costs contributes 15%. All the other ment perspective, WWTP operators are able to broadly be identi-
cost items represent a percentage equal to or lower than 10%. In ed in two groups of WWTPs, that is, efcient and inefcient
spite of the fact that the size of the WWTPs evaluated in this study facilities. WWTP managers should identify the characteristics and
is medium-large, their operating cost distribution is consistent with operational practices of the efcient plants, as they are the best
that reported by previous studies (Molinos-Senante et al., 2010; practices. Subsequently, these best operational practices should be
Sala-Garrido et al., 2012b). implemented in the inefcient WWTPs in order to improve their
efciency. Thus, the operational costs of the WWTPs would be
4. Results and discussion reduced, while maintaining the quality of the treated water.
Before estimating the efciency scores for each WWTP, it was The efciency score of each WWTP depends on the weights
necessary to assign the weights for each input and output involved attributed to each input and output and although weights were
in the WSBM model. Following the procedure described in Section calculated based on their shadow prices, they are subjected to some
2.2 and taking into account that the sum of all output and input uncertainty. In this context, the uncertainty level for energy weight
weights should be 100, Table 2 shows the average value of the input is greater than for outputs weights. This is because although ef-
weights and output weights for estimating the efciency of ciency assessment considered six inputs, only the energy item was
WWTPs. weighted according the shadow price of CO2 emissions associated
Table 2 illustrates signicant differences between the weights to energy consumption. Moreover, it should be noted that energy is
assigned to the outputs depending on their environmental impact. a resource whose price is highly variable. Therefore, to narrow the
It is illustrated that both nutrients, nitrogen, and phosphorus have uncertainty associated to the weight attributed to energy costs, a
weights considerably greater than the other pollutants. This means sensitivity analysis that considers variations in the weight assigned
that the negative environmental impact associated with the to energy was made. So that the WSBM model was run with dif-
discharge of these pollutants to the environment would lead to one ferences that are 10% 25% 50% 75% and 100% from the
of the main problems of water body contamination: eutrophication. initial energy weight. The energy weights are shown as supple-
Once the weights of each input and output were assigned, the mental material.
efciency score for each WWTP was computed by applying the Table 4 shows a summary of the efciency scores related to the
WSBM model introduced by Tone (2011). As it is shown in Table 3, 11 different weights attributed to energy input. It is illustrated that
the average efciency score obtained for all facilities is 0.5660, changes in energy weights do not affect signicantly to the average
indicating that on average WWTPs could reduce their costs by 43% of the efciency of the sample of WWTPs evaluated. It is evidenced
while obtaining treated water with the same quality as currently as well that the number of efcient plants is the same (20 out of 49)
produced. However, it should be considered that the efcient in the 11 scenarios evaluated. In order to verify whether these small
Table 1
Sample description.
Table 2
Weights established for inputs and outputs.
25.3381 14.9324 14.9324 14.9324 14.9324 14.9324 0.0095 0.1604 32.686 67.1441
Table 3
Summary of the efciency scores obtained with the WSBM model.
Number of WWTPs WWTPs (percentage) Average of efciency score Standard deviation of efciency Saving potential (percentage)
0.8
4.3. Potential economic saving
0.6
Because the WSBM model is a non-radial DEA approach it allows
0.4
the reduction in each input (cost item) that is necessary for each
0.2 WWTP to be efcient to be estimated. In other words, the WSBM
model allows potential economic savings in each cost item to be
0.0 quantied. This information is essential for WWTPs manager to
0 10 20 30 40 50
support the decision-making process. The quantication of the
WWTPs potential economic savings is fundamental to selecting the priority
Fig. 1. Efciency score by WWTP.
actions to be implemented in the WWTPs. Tables S4 and S5 in
supplemental material section show the potential economic sav-
ings that each plant could obtain if they were efcient. According to
50 average values, the items in which WWTPs could save the most
costs are staff and energy. Hence, the WWTP operators should focus
40
on implementing operational and managerial measures to optimize
% of WWTPs
Table 4
Summary of efciency scores considering different weights for energy input.
Energy Energy Energy Energy Energy Energy Energy Energy Energy Energy Energy
weight 100 weight 75 weight 50 weight 25 weight 10 weight 10 weight weight 25 weight 50 weight 75 weight 100
percentage percentage percentage percentage percentage percentage original percentage percentage percentage percentage
Average 0.573 0.564 0.566 0.568 0.571 0.572 0.584 0.574 0.576 0.578 0.581
Standard Deviation 0.370 0.378 0.376 0.374 0.372 0.371 0.361 0.369 0.368 0.366
0.364
Minimum 0.138 0.127 0.130 0.132 0.135 0.137 0.153 0.139 0.141 0.146 0.150
Maximum 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000 1.000
Number 20 20 20 20 20 20 20 20 20 20 20
efcient
WWTPs
changes in efciency when energy weights change occur not only 2012a). Hence, it is fundamental that WWTPs' managers optimize
in the average scores but also at WWTP level, Fig. 3 shows the the use of blowers, turbines, etc. according time hours.
variation levels (maximum and minimum values) of the efciency Moreover, Tables S4 and S5 show a noticeable variability among
scores for each WWTPs under the 11 scenarios evaluated. It is WWTPs. As was expected, a set of facilities (efcient WWTPs) have
veried that at plant level, efciency scores remain almost constant no potential to save costs. On the other hand, other plants (inef-
although the weights attributed to energy item change noticeably. cient WWTPs) have a large potential saving. It should be high-
This nding means that efciency scores of WWTPs do not depend lighted that the maximum potential saving is V0.78/m3 which
L. Castellet, M. Molinos-Senante / Journal of Environmental Management 167 (2016) 160e166 165
1.0
0.9
0.8
0.7
Eciency score
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49
WWTPs
Maximum Original Minimum
Fig. 3. Maximum, original and minimum efciency scores of the WWTPs under different energy weights.
corresponds to WWTP11. Taking into account the volume of the rst time the efciency of a sample of WWTPs by introducing
wastewater treated annually, the potential savings for inefcient technical, economic, and environmental issues. In doing so, the
WWTPs is not negligible (see supplemental material). If the 29 WSBM model was applied, which is a non-radial DEA model which
inefcient WWTPs were to become efcient, they could collectively allows weights to be assigned to the inputs and outputs according
save more than V22 million annually. As is shown in Fig. 4, inef- their importance. Moreover, this model allows the potential eco-
cient facilities could save around 38% of their personnel costs and nomic savings of each cost item to be quantied at the WWTP level.
25% of their energy costs. Moreover, they could save 18%, 9%, 8%, The results from a sample of Spanish WWTPs provide the
and 3% on waste management, maintenance, reagents, and other following primary conclusions: (i) shadow prices of pollutants are a
costs, respectively. good proxy to assign weights to outputs; (ii) there are no facilities
with a moderate efciency score, since WWTPs were identied as
5. Conclusions either efcient or as inefcient plants (with an efciency score
lower than 0.6); (iii) the efciency scores of the WWTPs do not
The efciency evaluation of WWTPs is becoming increasingly depend on the weight attributed to the energy costs; and (iv) the
important as it identies those plants that make better use of their largest potential economic savings are associated with staff and
economic resources without reducing the quality of the water they energy costs.
treat. Thus, companies can identify the best operational practices From a policy perspective, the methodology and empirical
that could be applied in other WWTPs to contribute to the reduc- application carried out in this study are of great interest for (waste)
tion of operational costs. Ultimately, and taking into account that water authorities and company managers. In many countries,
wastewater treatment services are paid for by citizens through wastewater treatment services are provided following a monopoly
water tariffs, the improvement of the efciency of WWTPs is also approach; therefore operators have no incentives towards ef-
benecial for society. ciency and innovation. In this context, benchmarking, assumes
From a methodological point of view, previous studies have strategic importance since through it, (waste)water authorities can
illustrated that DEA is a suitable technique for evaluating the ef- promote efciency in WWTPs which is essential to reduce the
ciency of WWTPs. According this approach, the operational costs of wastewater treatment tariffs paid by citizens. From WWTPs'
the WWTPs are their inputs, while the pollutants removed from the managers perspective, the assignment of weights to the different
wastewater are the outputs. The so-called techno-economic ef- pollutants allows environmental issues to be integrated in the ef-
ciency of WWTPs is evaluated in this way. However, these previous ciency assessment of WWTPs. It will create incentives to operators
studies have ignored the different environmental impacts of the to improve the quality of the efuent of WWTPs. Second, WWTP
pollutants removed from the wastewater if they were dumped into managers can compare the performance of WWTPs and therefore
water bodies. To overcome this limitation, this paper evaluates for identify best operational practices. The implementation of these
practices in inefcient plants would contribute to improving their
efciency and consequently to reducing operational costs. Third,
3% the quantication of the potential economic savings of each cost
25% item is essential to support the decision-making process. It pro-
18% vides information for the prioritization of the measures that are to
be implemented in the WWTPs to further reduce operational costs.
banks: non-radial directional performance measurement with undesirable 2011. Economic feasibility study for phosphorus recovery processes. Ambio 40,
output. Omega Int. J. manage. s. 40, 1e8. 408e416.
Carvalho, P., Marques, R.C., 2011. The inuence of the operational environment on Murillo-Zamorano, L.R., 2004. Economic efciency and frontier techniques. J. Econ.
the efciency of water utilities. J. Environ. Manag. 92, 2698e2707. Surv. 18, 33e45.
Cooper, W.W., Seiford, L.M., Zhu, J., 2011. Handbook on Data Envelopment Analysis. Ontiveros, G.A., Campanella, E.A., 2013. Environmental performance of biological
Springer, New York. nutrient removal processes from a life cycle perspective. Bioresour. Technol.
Gjebrea, E., Zoto, O., 2013. Regionalization of water supply and sewerage companies 150, 506e512.
as a solution for the efciency of water supply and sewerage sector: case of Papong, S., Rotwiroon, P., Chatchupong, T., Malakul, P., 2014. Life cycle energy and
Albania. Mediterr. J. Soc. Sci. 4, 37e45. environmental assessment of bio-CNG utilization from cassava starch waste-
Guerrini, A., Romano, G., Leardini, C., Martini, M., 2015. Measuring the efciency of water treatment plants in Thailand. Renew. Energ 65, 64e69.
wastewater services through data envelopment analysis. Water Sci. Technol. 71, Rodriguez-Garcia, G., Molinos-Senante, M., Hospido, A., Hern andez-Sancho, F.,
1845e1851. Moreira, M.T., Feijoo, G., 2011. Environmental and economic prole of six ty-
Herna ndez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2010. Economic pologies of wastewater treatment plants. Water Res. 45, 5997e6010.
valuation of environmental benets from wastewater treatment processes: an Sala-Garrido, R., Herna ndez-Sancho, F., Molinos-Senante, M., 2012b. Assessing the
empirical approach for Spain. Sci. Total Environ. 408, 953e957. efciency of wastewater treatment plants in an uncertain context: a DEA with
Herna ndez-Sancho, F., Molinos-Senante, M., Sala-Garrido, R., 2011. Energy efciency tolerances approach. Environ. Sci. Policy 18, 34e44.
in Spanish wastewater treatment plants: a non-radial DEA approach. Sci. Total Sala-Garrido, R., Molinos-Senante, M., Hern andez-Sancho, F., 2012a. How does
Environ. 409, 2693e2699. seasonality affect water reuse possibilities? an efciency and cost analysis.
Hoang, V., 2014. Analysis of resource efciency: a production frontier approach. Resour. Conserv. Recycl. 58, 125e131.
J. Environ. Manag. 137, 128e156. Tone, K., 2001. Slacks-based measure of efciency in data envelopment analysis.
Lannier, A.L., Porcher, S., 2014. Efciency in the public and private French water Eur. J. Oper. Res. 130, 498e509.
utilities: prospects for benchmarking. Appl. Econ. 46, 556e572. Tone, K., 2010. Variations on the theme of slacks-based measure of efciency in
Molinos-Senante, M., Hanley, N., Sala-Garrido, R., 2015b. Measuring the CO2 DEA. Eur. J. Oper. Res. 200, 901e907.
shadow price for wastewater treatment: a directional distance function Tone, K., 2011. Slacks-based measure of efciency. In: Cooper, W.W., Seiford, L.M.,
approach. Appl. Energy 144, 241e249. Zhu, J. (Eds.), Handbook on Data Envelopment Analysis. Springer, New York, NY,
Molinos-Senante, M., Herna ndez-Sancho, F., Sala-Garrido, R., 2010. Economic pp. 195e210.
feasibility study for wastewater treatment: a cost-benet analysis. Sci. Total Wang, X.H., Wang, X., Huppes, G., Heijungs, R., Ren, N.Q., 2015. Environmental
Environ. 408, 4396e4402. implications of increasingly stringent sewage discharge standards in municipal
Molinos-Senante, M., Hernandez-Sancho, F., Sala-Garrido, R., 2014. Benchmarking wastewater treatment plants: case study of a cool area of China. J. Clean. Prod.
in wastewater treatment plants: a tool to save operational costs. Clean. Technol. 94, 278e283.
Environ. 16, 149e161. Wu, J., Xu, D., He, F., He, J., Wu, Z., 2015. Comprehensive evaluation of substrates in
Molinos-Senante, M., Herna ndez-Sancho, F., Sala-Garrido, R., 2015a. Comparing the vertical-ow constructed wetlands for domestic wastewater treatment. Water
dynamic performance of wastewater treatment systems: a metafrontier Pract. Technol. 10, 625e632.
Malmquist productivity index approach. J. Environ. Manag. 161, 309e316. Zhu, J., 2015. Data Envelopment Analysis: a Handbook of Models and Methods.
Molinos-Senante, M., Herna ndez-Sancho, F., Sala-Garrido, R., Garrido-Baserba, M., Springer, New York.