Addressing agricultural pressures on water resources: a DEA environmental assessment in the case of european transboundary basins

169 169 DOI: 10.1201/9781003240129-9 Addressing Agricultural Pressures on Water Resources A DEA Environmental Assessment in the Case of European Transboundary Basins A. Expósito and F. Velasco 9.1 Introduction ...................................................................................................169 9.2 Case Study .....................................................................................................170 9.3 Methodology and Data ..................................................................................171 9.3.1 Method ..............................................................................................171 9.3.2 Data ...................................................................................................175 9.4 Results ...........................................................................................................177 9.5 Conclusion .....................................................................................................181 9.1 INTRODUCTION Agriculture is not only the main user of water resources (around 70 percent of all water abstraction worldwide) (FAO 2017) but is also the most intractable source of riverine and marine pollution (Boyle 2014; Mohaupt et al. 2007). Most European river basins (RBs) face significant pressures from agriculture, both in terms of quantity (e.g., abstraction stress) and quality (e.g., agrochemical diffuse pollution) (EEA 2015; EC 2012; Mohaupt et al. 2007; Özerol et al. 2012). The European Union (EU) water legislation aims to achieve an ecologically and chemically good status of water bodies through the reduction of demand pressures on water resources, a progressive reduction of pollution, and the preservation of ecosystems (van Rijswick et al. 2010). The European Environment Agency (EEA) (EEA 2012a, 2015) indicates that the environmental status of a large number of water bodies across Europe is unacceptable and is of greater concern in large transboundary river basins (TRBs). Furthermore, the EEA asserts that emissions of nutrient (e.g., phosphorus and nitrogen) compounds from agriculture represent a serious pollutant threat to European water bodies (EEA 9 9780367758110_pi-214.indd 1699780367758110_pi-214.indd 169 19-Aug-21 21:06:1619-Aug-21 21:06:16 170 Renewable for Climate Mitigation 170 2012b; EU 2017). The EU Water Framework Directive (WFD) (EC/ 2000/ 60) explicitly includes transboundary rivers, for whose management international or interregional structures are envisaged (Art. 3, Preamble 35). However, this directive fails to specify instruments for coordination and responsibilities for riparian states so that the established environmental goals may be more easily achieved (van Rijswick et al. 2010; Wiering et al. 2010). Transboundary rivers represent complex interdependent structures at a multidimensional level (e.g., socioeconomic, political, environmental) for riparian countries, whereby the river is a source of externalities (e.g., agrochemical pollution) and mutual vulnerabilities (Dimitriou et al. 2012; Dinar et al. 2013; Kauffman 2015). In this respect, the subtractability and nonexclusion characteristics of transboundary rivers have led them to become common pool resources shared by riparian countries where the common dilemma of overuse and mismanagement can easily occur (Ostrom 2005). Much of the recent research on the management of transboundary rivers has focused on the determinants of political and geographical conditions for effective cooperative management (e.g., Dinar et al. 2013; Zawahri and Mitchell 2011), legal aspects and conflict resolution (e.g., Jager 2016; Petersen- Perlman et al. 2017), and risks associated to climate change, such as droughts and floods (e.g., Bakker and Duncan 2017; Pulwarty and Maia 2015). Although environmental degradation is considered an incentive for international joint efforts (Dinar et al. 2013) and agricultural pressures on water resources represent a significant challenge for RB management (Kallioras et al. 2006), their analysis has attracted much less attention in academic and international cooperation spheres (Munia et al. 2016; UNECE 2011). The present work aims to fill this gap and offers a comparative analysis for the identification of factors that may influence performance patterns in managing agricultural pressures. With this aim, the data envelopment analysis (DEA) methodology used in this study enables the assessment of the capacity of the analyzed RBs regarding the minimization of agricultural pressures on water resources. The rest of the paper is organized as follows. Section 9.2 briefly describes our case study comprised of 20 European RBs. The following section (Section 9.3) presents the DEA methodology used and describes the data. Section 9.4 summarizes the results of our analysis. Finally, certain concluding remarks are offered in Section 9.5. 9.2 CASE STUDY TRBs comprise about 47 percent of the world’s continental land area and Europe has the largest number of these transnational rivers, that is, 68 out of the total 286 in the world (Giordano et al. 2014). The watershed or RB scale has often been considered a useful organizational unit for the evaluation of policy and research issues, especially regarding environmental concerns (EC 2012). Further, the WFD recognizes the RB as the main natural unit for the protection of the status of water bodies, and as the appropriate scale for integrated water resource management (IWRM) (Berbel and Expósito 2018). In this respect, the DEA method applied herein, which is explained in greater detail in the subsequent section, uses RBs as the decision units to be assessed. Specifically, the method applied in this study allows us to assess the management 9780367758110_pi-214.indd 1709780367758110_pi-214.indd 170 19-Aug-21 21:06:1619-Aug-21 21:06:16 171 Agricultural Water Resources 171 performance of these decision units (or RBs in our case study) at reducing agricultural pressures, such as nutrient pollution and abstraction stress, on water resources. Over recent decades, cooperation efforts, involving all TRBs to solve environmental problems derived from human activities (including agriculture), have increased dramatically, as globally initiated by the Global Environment Facility (GEF) set up by the World Bank in 1991 (Gerlak 2004), and followed by the Convention on the Protection and Use of Transboundary Watercourses and International Lakes (UNECE, Helsinki, 1992; Libert 2015). To this end, the Convention established the principle of the joint management bodies, based on any multilateral institutional arrangement for cooperation between the riparian countries. Since then, most European TRBs have created international commissions to jointly manage water resources on a basin scale and to reduce negative externalities from human activities, such as agricultural pollution and abstraction stress. One of the first multilateral cooperation institutions was established in 1988 through the International Commission for the Protection of the Danube river (ICPDR), whose main objective is the implementation of the Danube River Protection Convention. Similarly, international commissions have been established for many other rivers in Europe, such as the Conventions of the International Commission for the Protection of the Elbe (1991), the Oder (1999), the Rhine (1999), and the Albufeira Convention for the Iberian rivers (1998). As further discussed in the “Discussion” section of this study, these transboundary commissions have been widely criticized for failing to have achieved effective joint management aimed at reducing pressures on water resources (including those of the agricultural sector), and hence for failing to have served as an effective instrument to improve the ecological and chemical status of TRBs (Bernauer and Kuhn 2010). In this context, this study takes a representative sample of 20 European RBs to assess their management capability in terms of reducing agricultural pressures on water resources. It is worth noting that the area of some RBs, such as Ebro and Po, are mainly located in one country. This study aims to assess the managerial efficiency at RB scale, thus taking the RB as a management decisionmaking unit, regardless of the number of riparian countries or the existence of TRB agency. Nevertheless, the potential effect of these factors, among others, on the estimated efficiency patterns are also analyzed in this study. Table 9.1 shows various characteristics regarding drainage area, population, per capita gross domestic product (GDP), renewable water sources (RWS), total water withdrawals, and the relative weight of agriculture on those total withdrawals. These selected indicators show the high heterogeneity in our sample of RBs. A further description of the riparian countries of each TRB is given in Table 9.2. 9.3 METHODOLOGY AND DATA 9.3.1 METHOD The DEA method was initially proposed by Farrell (1957). Subsequently, it has been extended under various functional schemes, such as an inputoriented scheme with constant returns to scale (Charnes et al. 1978), outputoriented maximization (Charnes et al. 1981), variable returns to scale (Banker et al. 1984), and both radial 9780367758110_pi-214.indd 1719780367758110_pi-214.indd 171 19-Aug-21 21:06:1619-Aug-21 21:06:16 172 Renewable for Climate Mitigation 172 and nonradial approaches (Sueyoshi and Sekitani 2009), among others. In the water management sector, DEA methods have been extensively used to assess efficiency among a group of management or decisionmaking units (DMUs) (e.g., water utilities) (Xiang et al. 2016; Romano et al. 2017, among others). Zhu (2016) offers a recent review of DEA literature and applications in environmental issues. DEA methods are based on a nonparametric approach to estimate efficient frontiers in the sense that no assumption regarding the functional form is required. This enables relative efficiency estimates to be obtained for a group of DMUs (i.e., our sample of European TRBs) by using multiple inputs and outputs and alternative output– input specifications. Furthermore, DEA methods measure the efficiency of a DMU with the simple restriction that all sampled DMUs lie on or below the efficient frontier and obviate the need to assign prespecified weights to either inputs or outputs. Each DMU not on the frontier (thus, an inefficient DMU) is scaled against a convex combination of the DMUs on the frontier faced closest to it. Thus, efficiency mappings of a group of DMUs can be obtained and efficient (or benchmark) DMUs can be identified. Additionally, and conversely to other methodologies, such as qualitative analysis and multicriteria schemes, DEA methods are capable of identifying TABLE 9.1 Indicators of Selected European TRBs (2010) River basin Drainage area (103 km2) Population (103) GDP pc (USD) RWS (km3/ year) Water withdrawals (km3/ year) Agric. withdrawals (% total) Danube 796 80,185 18,478 221,762 53,822 36.0 Dnieper 511 29,457 5,889 66,635 14,486 35.5 Don 439 18,819 11,359 45,375 10,205 29.4 Douro/ Duero 97 3,492 25,521 24,098 7,412 79.5 Ebro 85 2,805 28,024 19,082 9,865 72.3 Elbe 139 21,860 37,940 28,957 7,462 9.8 Guadiana 67 1,475 28,017 11,076 8,605 94.0 Kemi 54 105 44,504 17,900 30 2.7 Klarälven 50 901 60,573 20,564 602 12.7 Maritsa 53 3,476 8,965 11,970 6,404 52.6 Neman 93 4,789 12,144 20,754 880 12.1 Oder/ Odra 119 15,718 15,163 20,997 4,720 4.1 Po 72 15,918 35,500 48,958 18,575 40.5 Rhine 164 48,831 49,543 74,972 28,834 4.5 Rhone 97 10,055 46,047 52,339 8,207 25.1 Seine 73 15,775 41,422 20,712 8,353 20.6 Tagus/ Tejo 71 7,244 28,303 19,297 7,968 59.9 Vistula/ Wista 192 23,148 12,753 34,604 7,699 3.4 Volga 1,412 58,621 14,612 274,165 25,004 11.4 Vuoksi 287 3,246 23,001 87,344 5,590 1.3 Source: Data from TWAP (2019). 9780367758110_pi-214.indd 1729780367758110_pi-214.indd 172 19-Aug-21 21:06:1619-Aug-21 21:06:16 173 Agricultural Water Resources 173 input– output relationships that remain unobserved for other methods. DEA methods therefore constitute a powerful instrument in obtaining additional strategic information not provided by alternative methodologies. DEA methods allow for two types of model orientations: input or output (Avkiran and Rowlands 2008). Since our objective is to estimate the relative efficiency of each DMU at generating selected outputs (i.e., reduction of agricultural pressures), we believe the output specification to be the most appropriate. The specific approach used in this study also accounts for the role of undesirable outputs, such as agricultural pressures (quantitative and qualitative) on water resources, in order to assess efficiency among a group of DMUs (e.g., TRBs). Recent DEA applications have shown the relevance of including undesirable outputs (e.g., pollution) as a more realistic specification of the optimization model (Sueyoshi and Goto 2011). Furthermore, recent DEA developments have revealed the importance of using different output specifications (i.e., natural and managerial) in order to obtain additional strategic information for the assessment of DMUs (Expósito and Velasco 2018; Sueyoshi and Goto 2011). With the aim to assess the managerial capacity of RBs at minimizing agricultural pressures, this study uses a managerial output specification. TABLE 9.2 Riparian Countries TRB Riparian countries Danube Albania, Austria, Bosnia And Herzegovina, Bulgaria, Croatia, Czech Republic, Germany, Hungary, Italy, (The former Yugoslav Republic of) Macedonia, (Republic of) Moldova, Montenegro, Poland, Romania, Serbia, Slovakia, Slovenia, Switzerland, Ukraine Dnieper Belarus, Russian Federation, Ukraine Don Russian Federation, Ukraine Douro/ Duero Portugal, Spain Ebro Andorra, France, Spain Elbe Austria, Czech Republic, Germany, Poland Guadiana Portugal, Spain Kemi Finland, Norway, Russian Federation Klarälven Norway, Sweden Maritsa Bulgaria, Greece, Turkey Neman Belarus, Latvia, Lithuania, Poland, Russian Federation Oder/ Odra Czech Republic, Germany, Poland, Slovakia Po France, Italy, Switzerland Rhine Austria, Belgium, France, Germany, Italy, Liechtenstein, Luxembourg, Netherlands, Switzerland Rhone France, Italy, Switzerland Seine Belgium, France Tagus/ Tejo Portugal, Spain Vistula/ Wista Belarus, Czech Republic, Poland, Slovakia, Ukraine Volga Kazakhstan, Russian Federation Vuoksi Belarus, Finland, Russian Federation 9780367758110_pi-214.indd 1739780367758110_pi-214.indd 173 19-Aug-21 21:06:1619-Aug-21 21:06:16 174 Renewable for Climate Mitigation 174 The managerial specification implies a high methodological sophistication in the sense that a DMU may decrease the directional vector of undesirable outputs by increasing inputs. This type of specification is referred to as “managerial disposability” and is usually related to the implementation of innovative management initiatives (e.g., effective multilateral arrangements to reduce agricultural pressures on water resources at TRB scale). In our specific case, it would reflect the TRB capacity to minimize agricultural pressures on water resources despite a potential increase of agricultural water withdrawals. The optimization model produces an autonomous indicator of relative efficiency referred to herein as “managerial efficiency” scheme. Subsequently, a simplified mathematical description of the applied DEA model is offered. Nevertheless, a detailed description of the DEA model applied can be found in Expósito and Velasco (2018, 2020). In our DEA model, each jth DMU jn=…1, , uses inputs Xx x jjmj =… () 1,, T and generates desirable outputs, represented by ,,Gg g jjsj =… () 1 T, and undesirable outputs, represented by Bb b jjhj =… () 1,, T. Furthermore, dd i x r g im rs ,,,, ,, , =… =…11 and df b fh,, , =… 1 represent slack variables related to inputs, and desirable and undesirable outputs, respectively. λλ λ =… () 1,, n T are unknown structural or intensity variables, which are used for connecting the input and output vectors via a convex combination. R is the range resolute throughout the upper and lower bounds of inputs, desirable outputs, and undesirable outputs and is expressed by following expressions: Rmsh xj nx jn x ii ji j ()max| ,, -min |,, - =++=… {} =… {} ( 1 11 )) -1 Rmsh gj ng jn g rr jr j ()max| ,, -min |,, - =++=… {} =… {} ( 1 11 )) -1 Rmsh bj nb jn b ff jf j ()max| ,, -min |,, - =++=… {} =… {} ( 1 11 )) -1 The managerial efficiency of the kth DMU is evaluated by the following radial model: Max ξε +++ ⎡ ⎣ ⎢ ⎢ ⎤ ⎦ ⎥ ⎥ === ∑∑∑ i m i x i x r s r g r g f h f b f b Rd Rd Rd 111 st xdxi m j n ij j o i x ik .. ,, , = ∑+− () ==… () 1 11 λ 9780367758110_pi-214.indd 1749780367758110_pi-214.indd 174 19-Aug-21 21:06:1719-Aug-21 21:06:17 175 Agricultural Water Resources 175 ,,, j n rj jr g rk rk gdgg rs = ∑−− ==… () 1 1 λξ ,,, j n fj jf b fk fk bdbb fh = ∑++ ==… () 1 1 λξ j n j = ∑= 1 1 λ , λ ji x jndi m,, ,,,, ≥=… () ≥=… () 01 01 dr sd fh r g f b ,,,≥=… () = () 01 01,… ,a nd, ≥ ξ :unrestricted (9.1) Its solution provides the necessary efficiency scores, measured by: θξ ε ** =1 i x* =1 * =1 * =1 +d++−⎛ ⎝ ⎜⎞ ⎠ ⎟ ⎡ ⎣ ⎢ ⎢ ⎤ ∑∑∑ i m i x r s r g r g f h f b f b RRdRd ⎦⎦ ⎥ ⎥, (9.2) being o=1 . Once the efficiency mapping of DMUs is estimated, this study tests the impact of certain determinants on the estimated efficiency scores for our sample of European TRBs. The most commonly included determinants (or explanatory factors) in the literature regarding the comparative analysis of the environmental performance of TRBs (see, for example, Bernauer and Kuhn (2010) and Knieper and Pahl- Wostl (2016), among others) are those related with population (e.g., population density), economic development (e.g., per capita income or GDP), overall human pressure on water resources (e.g., water withdrawals on renewable water resources on a RB scale), and geolocation variables. In order to test whether these determinants carry a sufficiently large influence to explain efficiency scores in our case study, we have applied both an ordinary least squares (OLS) model, which controls for heteroscedasticity with robust standard errors, and a Tobit model, which controls for upper limit values (i.e., 1 = full efficiency). The use of two alternative regression models contribute to the robustness of the obtained results. 9.3.2 DATA In 2012, the GEF approved the Transboundary Water Assessment Programme (TWAP) of the United Nations Environment Programme (UNEP) following an 9780367758110_pi-214.indd 1759780367758110_pi-214.indd 175 19-Aug-21 21:06:1719-Aug-21 21:06:17 176 Renewable for Climate Mitigation 176 earlier programme (Development of the Methodology and Arrangements for the GEF Transboundary Waters Assessment Programme in 2011). TWAP offers a unique homogeneous dataset for TRBs that enables international comparison and assessment studies (UNEP 2016). TWAP constitutes a global assessment platform of 286 TRBs distributed worldwide and offers a wide variety of indicators regarding water quantity, water quality, ecosystems, governance, and socioeconomic fields. This study uses the latest available data (year 2010) of the TWAP dataset (TWAP, 2019). In order to apply our DEA assessment to the 20 European TRBs selected, the following indicators have been employed: 1. Agricultural withdrawals (input indicator): This indicator has been selected as the necessary input for agricultural activities (i.e., irrigation and livestock production). This input is withdrawn from the RB system (both surface and groundwater sources) for production objectives and given back (partially) in the form of returns which usually contain pollutants, such as nitrogen and phosphorus compounds. Values are measured in km3/ year (Table 9.3). 2. Nutrient pollution (qualitative output indicator): This indicator considers river pollution from nitrogen and phosphorus compounds, which are mainly caused by agricultural activities (Bernauer and Kuhn 2010; Dimitriou et al. 2012; EEA 2012b). Urban wastewater and atmospheric deposition of nitrogen can also generate nutrient pollution, but to a much lesser extent (Boyle 2014). The construction of the indicator is based on the methodology developed by Mayorga et al. (2010) and Seitzinger et al. (2010). In our specific DEA optimization model, this indicator needs to be minimized. Values range from 0 to 1 (from less to more highly polluted) (Table 9.3). 3. Agricultural abstraction stress (quantitative output indicator): This indicator identifies agricultural water stress, and is constructed based on the consumption- to- availability ratio (mean annual water consumption divided by the sum of mean annual runoff on a river basin scale). A reduction of the agricultural stress indicator implies a decrease in the pressure of agriculture on water resources of the RB (e.g., through more efficient practices of water use). Its introduction as an additional output indicator is justified since the presence of agricultural nutrient pollution in RBs is strongly related to quantity factors, as measured by consumption- to- availability ratios (EEA 2012b; TABLE 9.3 Descriptive Statistics Agricultural withdrawals Nutrient pollution Abstraction stress Average 3,688 0.69 0.06 St. Deviation 4,589 0.21 0.07 Maximum 19,397 1 0.30 Minimum 0.83 0.25 0.00 9780367758110_pi-214.indd 1769780367758110_pi-214.indd 176 19-Aug-21 21:06:1719-Aug-21 21:06:17 177 Agricultural Water Resources 177 Munia et al. 2016). In our specific case, this output has been introduced in terms of its inverse value, therefore this indicator needs to be maximized in our optimization problem. Values range from 0 to 1 (Table 9.3). In summary, our optimization problem aims to minimize not only the agricultural nutrient pollution (qualitative output indicator) but also the agricultural abstraction stress (quantitative output indicator). Therefore, the DEA method applied enables both output indicators to be simultaneously optimized. Descriptive statistics of our input and output variables are shown in Table 9.3. It is worth noting that the contrast of output/ input ratios of each RB with the mean values shows that any potential problems of sensitivity caused by the existence of outliers and statistical noise are ruled out. DEA methods do not require data to be normalized (e.g., by agricultural activity or withdrawals in the basin) to obtain efficiency scores. Additionally, the condition regarding the minimum number of observations per variable established by Banker et al. (1996) is met (20 DMUs and 3 input– output variables). Hence no misspecification problems are observed. 9.4 RESULTS As discussed in the previous section, relative efficiency relies on effective management initiatives to minimize undesirable outputs. In our managerial output specification, the imposition of the reduction of undesirable outputs can be achieved despite an input increase (i.e., higher water withdrawals). Under this assumption, the obtained efficiency mapping describes the capacity to reduce agricultural pressures on water resources (despite potential higher withdrawals of water for agriculture) and helps to identify those benchmark RBs from which to learn. The results of our efficiency assessment for our sample of European RBs are presented in a stepby- step manner. The first step focuses on the efficiency mapping obtained from our model specification (Table 9.4). The results identify four efficient (benchmark) RBs: the Danube, Volga, Kemi, and Vuoksi RBs. In these RBs, an increase in agricultural withdrawals would not necessarily imply greater pressures on water resources from agriculture. This can be explained by the implementation of effective management initiatives that positively impact on the input– output intrinsic relationship. Conversely, inefficient RBs would therefore need to increase their management efforts toward reducing agricultural pressures, since the increase of agricultural withdrawals would be counterproductive. Within the heterogeneous range of inefficient RBs, the Elbe and Rhine RBs register the lowest efficiency scores (0.25), thus showing the urgent need to implement agricultural management initiatives at basin scale. Interestingly, the Kemi and Vuoksi RBs, two efficient DMUs, present a very low proportion of agricultural water withdrawals on total withdrawals on a basin scale, which could be argued as a possible explanatory factor for its high efficiency scores (since lower agricultural water withdrawals would lead to less agricultural nutrient pollution and less abstraction stress). Nevertheless, the Volga and Danube RBs register greater agricultural water withdrawals (Table 9.1) and are also fully efficient DMUs in our managerial specification. Therefore, as widely argued in the existing literature (Zhu 2016), DEA outcomes do not depend on input data (or on its 9780367758110_pi-214.indd 1779780367758110_pi-214.indd 177 19-Aug-21 21:06:1719-Aug-21 21:06:17 184 Renewable for Climate Mitigation 184 Zawahri NA, Mitchell SM (2011) Fragmented governance of international rivers: Negotiating bilateral versus multilateral treaties. Int Stud Quart 55(3):835– 858. Zhu J (2016) Data Envelopment Analysis: A Handbook of Empirical Studies and Applications. New York: Springer. 9780367758110_pi-214.indd 1849780367758110_pi-214.indd 184 19-Aug-21 21:06:1719-Aug-21 21:06:17