Efficient water management: an analysis for the agricultural sector

Water shortage has become a concern for several stakeholders related to water use and management, from the policymakers to the end consumer. However, sometimes, it is not easy to implement any strategy to save freshwater, either due to the perceptions of the different stakeholders, or the lack of adjusted plans and strategies to improve efficiency (Marston & Cai, 2016).

Specifically, in the agricultural sector, these questions related to efficient freshwater management hold special relevance, not only for environmental and sustainability problems, but, also, in reducing farmers’ costs. Public institutions and policymakers have a determinant role to play here, namely, in the framework of the Common Agricultural Policy (CAP) inside the European Union (Salmoral et al., 2017).

In fact, across the world, water withdrawal by the agricultural sector (around 2010) was about 70%. This question assumes special relevance in Africa and Asia (81%) and Oceania (65%). In European and American countries, the situation is not so problematic (25% and 48%, respectively), but deserves also special attention, namely, in the southern countries (Aquastat, 2019).

To increase the effectiveness of the implementation of any plan or strategy to promote efficiency in water use on farms, it is fundamental to involve several stakeholders, namely, farmers, in the design and implementation process. This will improve farmers’ perceptions and will allow for adjustments based on the stakeholders’ experiences, problems and concerns (Esteban et al., 2018).

In this way, it is crucial to bring more insight regarding these aspects associated with efficient water management in the agricultural sector of the European Union regions and countries. In this context, the main objective of the research presented here is to analyse the efficiency of water management by the agricultural sector within the European Union. For the literature survey, mainly the whole databases from the Web of Science (2019) were considered. To organize the literature review the Atlas.ti (2019) software was taken into account. Following this, an analysis of the data, obtained from Eurostat (2019), was carried out (other databases could be considered, however the information available in the Eurostat seems to be more adjusted to the objectives proposed in this research), and this statistical information was explored through an approach that combined factor and cluster analysis with data envelopment analysis (DEA). The factor and cluster analysis was considered to find more homogeneous decision making units (DMUs). With more homogeneous groups of European Union regions and countries, the intention was to reduce the influence, in the efficiency analysis, from national and regional specific characteristics. Before the cluster analysis, factor analysis was used to avoid problems of collinearity between the variables. After the factor and cluster analysis, considering Stata (2019) and Torres-Reyna (n.d.) procedures, the DEA through the DEAP (2019) software was performed. For the DEA the Cobb & Douglas (1928) model from the theory of production was considered as the base. This research follows studies such as, for example, Ali & Klein (2014), Xiang et al. (2016), Martinho (2017) and Suárez-Varela et al. (2017). These authors considered the DEA methodologies so as to analyse the issues related to the sustainable use of resources, including water. Approaches combining qualitative analysis, with factor-cluster analysis and DEA seem to be a field where there are issues to be explored, namely, those related to water efficiency.

This study delivers a systematic analysis about the literature related to efficient water management in the agricultural sector and highlights the potentialities of water savings in European countries and regions. These are relevant insights for the several operators related to water use and management (farmers and policymakers), based on new approaches that combine qualitative analysis (considering the Atlas.ti software) with the DEA–factor-cluster analysis.

To better organize the literature analysis, first a qualitative analysis will be done considering the Atlas.ti software and 50 of the 54 scientific documents obtained from the Web of Science databases related to the topics: ‘efficient water management’; agriculture. Only 50 were considered because the Atlas.ti was unable to import information for the following studies: Cassel et al. (1978), Barrett & Skogerboe (1980), Rahman et al. (1981) and Hundal & Dedatta (1984).

Figure 1 presents the main words in the 50 scientific documents related to efficient water management by the agricultural sector. It is proposed that the following three groups can be considered: irrigation efficiency, water shortage, water quantification. In the following subsections, these three groups will be explored in depth, through a literature survey, and several studies are considered here.

Efficient water management in irrigation practices depends on several factors (Koscielniak et al., 2005), as those related to the political, economic, social, technological, legal and environmental contexts (Nazari et al., 2018). Among these factors, the depth of application, for instance, may have relevance in these contexts, as well as adjusted drainage systems (Chandio et al., 2013). In turn, the irrigation practices are interrelated with other agricultural techniques, such as soil management approaches (Rahman & Islam, 1991; Lee et al., 2018) or energy consumption (Yuan et al., 2018). In fact, the way in which irrigation is performed can have a significant impact on water management efficiency, namely, in surface irrigation (Castanedo et al., 2013), on agricultural production (Ismail et al., 2008) and soil productivities (Kim et al., 2018). It also has implications on farms’ economic indicators (Kumar et al., 2009) and on the morphology and spatial distribution of the roots from permanent crops (Deng et al., 2017).

In the framework of the European Union requirements (e.g., WFD, 2000) for efficient water management, the water pricing systems, namely, those which charge farmers for the costs produced, in some circumstances may bring negative environmental externalities, such as putting more pressure on the demand for alternative water sources (Dono et al., 2010). These findings show the importance of monitoring and assessing the several policy instruments before and after their application (Kahil et al., 2015), to better achieve the objectives proposed and to be able to choose the best adjusted strategies (Kumar & Singh, 2003).

Farmers’ involvement in the whole design process of policy and water management plans improves the efficacy of their implementation. Another important aspect is to support farmers in choosing the best practices to improve the efficiency in water management with benefits for both the environment and the farmers’ income (Levidow et al., 2014).

In any case, water management efficiency has significantly improved around the world, thanks to developments in the technologies and techniques of irrigation or due to farmers’ options and decisions being more adjusted (Roth et al., 2013); nevertheless, it is still important to improve upon this (van Steenbergen et al., 2015).

Water scarcity is and will be one of the biggest challenges for the several stakeholders concerned with sustainable development, namely, for the water managers and policymakers. Considering water scarcity and its importance as a resource for the agricultural sector (Zhang et al., 2014a), the reuse of treated water may be an interesting approach towards dealing with this environmental and sustainability problem (Alcon et al., 2010), namely, reutilizing water from food industry effluent (Duek, 2016). The shortage of water will increase in the future because of climate change and global warming, with significant impacts on the percentage of irrigated land (D'Agostino et al., 2014) and promoting persistent dry seasons (Jin et al., 2018). The global increased demand for food and bioenergy sources also puts pressure on the availability of water (Graveline & Merel, 2014), such as with soil and water pollution (Muttamara & Sales, 1994).

Another relevant way to deal with the water shortage is to design an optimal allocation of irrigation to each stage of crop evolution, reducing hydric stresses and improving farmers’ profit (Fatemi et al., 2011). Another such aspect concerns the choice of cultivars (turf industry) having a lower need for water (Githinji et al., 2009). Efficient water management plans here make a determinant contribution.

Viticulture and wine production is a particular sector where water scarcity brings and will bring serious challenges, namely, in maintaining the specific characteristics of the certified wine, and where it is crucial for adjusted hydric stress management (Lanari et al., 2014). Rice cultivation is another agricultural production where water scarcity raises special concerns (Mostafa & Fujimoto, 2014) and where it is fundamental to implement efficient water management practices (Sharma & Rajput, 1990), such as mulching, for example (Totin et al., 2013). Mulching practice also has its application and use in other agricultural sectors, namely, when combined with surface irrigation (Zhang et al., 2014b), because, indeed, there are other farming sectors that may be affected by the water shortage, as, for example, maize, vegetable and fruit crops (including dry fruits).

This subsection intends to highlight methodologies, in general, considered to quantify the water availability and to present approaches that, in some contexts, may be considered as alternatives to those explored in this study. For example, to understand the water dynamics, it is important to analyse subsurface water movement (Shaw et al., 2001) and the relationships between the various water reservoirs, which may highlight important insights for the several stakeholders and it is here where methodologies such as oxygen-18 (180) and deuterium isotopes may produce relevant outcomes (Aly et al., 1993). Another interesting approach is the surface energy balance algorithm for land (SEBAL) for evapotranspiration analysis (Atasever & Ozkan, 2018), or METRIC (mapping evapotranspiration at high resolution using internalized calibration) (Chavez et al., 2009), or nonparametric approaches such as random regression forest models (RRFM) (Gonzalo-Martin et al., 2017), wireless in-field sensing and control (WISC) software (Kim et al., 2008; Kim & Evans, 2009), decision support systems for canal water release (CWREDSS) (Rao & Rajput, 2009), or surface renewal (SR) methodology (Rosa & Tanny, 2015).

Other important questions relate to the calculation of water requirements for plants (or the levels of evapotranspiration) and the availability of water, namely, in the soil (Wijewardana & Galagedara, 2010). Here, some indicators/indices such as the crop coefficient, or field capacity (Constanza Daza-Torres et al., 2017), or the leaf area index (Gassmann et al., 2011), or trunk sap flow (Liu et al., 2009) may be useful for the design of adjusted plans for water management (Haofang et al., 2017).

In any case, each approach needs to be validated for its location relative to specific conditions of each context (Otero et al., 2012) and farming production (Yang et al., 2016), namely, in the decision support systems, to avoid bias results and rejections by the several stakeholders related to water management (Kinzli et al., 2015).

To design better water management practices, it is important to quantify evapotranspiration and the water requirements of plants and it is here where new information technologies and communication and mathematical models may provide useful support (Kumar et al., 2012; Kumar, 2013). The real-time assessment of soil quality and water requirements and automated irrigation approaches with multi-sensors has gained increasing importance over the last few years (Miller et al., 2014).

Unfortunately, data about water use/withdrawal are not abundant in the international mainstream databases and less specifically for the agricultural sector. In any case, it was possible to obtain statistical information from the Eurostat for some European Union regions and countries, presented in Tables 1–3, with a robust time series for the period 2010–2012. Considering the approaches used here (the DEA), averages over this period for cross-section analysis were obtained.

The countries presented in Table 1 were selected due to the availability of data and this statistical information shows that Spain is where more fresh surface and groundwater is withdrawn for irrigation in the agricultural sector (about 22,394 million cubic metres) followed by France (about 10% of that of Spain), next Bulgaria (796.11 million cubic metres) and Romania (310.33 million cubic metres).

On the other hand, a great part of the water withdrawal was surface water in the majority of the European Union countries considered, except for Denmark, Cyprus, Lithuania, Malta, the Netherlands (where there is the inverse), France and the United Kingdom (where the values are almost the same).

At national and regional levels (NUTS 2), Tables 2 and 3 show a similar pattern to that presented in Table 1 in the relationships between surface and groundwater withdrawal. However, it is important to stress the great diversity of relationships for the French regions (the database does not present values for France, as a country, in the data disaggregated at regional level) when it is compared for surface and groundwater withdrawal. For example, in the Centre (French region) the greatest volume of water withdrawal was from groundwater (251.75 million cubic metres from groundwater and 28.20 million cubic metres from surface) and the inverse happens for Provence-Alpes-Côte d'Azur (626.04 million cubic metres from surface and 38.14 million cubic metres from groundwater).

In turn, within each country, there is also a diversity of realities, showing that the contexts between regions of the same country, with regards to water withdrawal for agricultural irrigation are different and the cluster analysis makes good sense.

For example, concerning water withdrawal from the surface (Table 2), the values for the region Yuzhen Tsentralen in Bulgaria represent almost 65% of the total and for the region Severozapaden only 0.3%. The same may be verified for the French context, where Provence-Alpes-Côte d'Azur reveals the highest values. The same disparities occur for the Hungarian and Polish frameworks.

For water withdrawal from groundwater (Table 3), the differences in the values between European Union regions, within each country, should also be taken into account in the analyses performed in the next section. For example, among the French regions, Centre and Aquitaine withdrawal alone accounts for almost half of the used water for agricultural irrigation.

To better organize the research presented here, namely, so as to avoid presenting too many tables, and considering the availability of data, in this and the next section, only statistical information for fresh surface water at national and regional level and for the year of 2010 will be explored. Fresh groundwater also is important in some European Union countries, as stressed before, but this may be considered in a future study considering the approaches covered here.

To improve understanding about the results presented in this section, Figure 2 shows the evolution of the standard output, utilized agricultural area and labour force across the European Union countries explored in this part of the study. This figure reveals that France and Poland, among the countries considered, are the member-states with higher standard output and utilized agricultural area; however, it stresses the comparatively (considering the values for the standard output and utilized agricultural area) low values for the labour force in France.

The factor and cluster analysis were performed following the Stata (2019) and Torres-Reyna (n.d.) procedures. To avoid problems of multicollinearity, before the cluster analysis the variables through factor investigation were also explored. For factor analysis, variables such as the standard output, utilized agricultural area, directly employed labour force and fresh surface water were considered. For the consideration of these variables, the Cobb–Douglas theory of production developments was taken into account as well as the design of the model for the DEA approach in the next subsequent section, for example, Martinho (2017).

By performing factor analysis the two factors presented in Table 4 were obtained. The main objective of the factor analysis is to find variables (factors) without having potential problems of multicollinearity to be considered for the cluster analysis. In Table 4 ‘variance’ is the total variance in the data explained by each factor, ‘difference’ is the difference between one variance and the next, ‘proportion’ is the per cent of total variance explained by each factor and ‘cumulative’ is the amount of variance explained by the current and all previous factors. In Table 5, the values for factor1 and factor2 show the relevance of each variable for the respective factor. For example, the correlation between the variable standard output for factor1 is 0.937 (values close to 1 represent a higher correlation). The uniqueness reveals the variance of the variable not related with the factors. It is expected to find lower uniqueness (Torres-Reyna, n.d.). In Table 6, the KMO (Kaiser-Meyer-Olkin) test is presented, and for this test, values of at least higher than 0.5 are expected (Filho & da Silva Júnior, 2010).

Table 4 shows that factor1 explains 99% of the total variance verified and Table 5 reveals that this factor1 is mainly defined by the standard output, utilized agricultural area and labour force. The fresh surface water little defines factor1 and the value for the uniqueness of this variable confirms the lower relevance of it for the explanation of the factors obtained. The sampling adequacy analysed in Table 6 also confirms the lower relevance of the fresh surface water for the factor analysis. In any case, the overall Kaiser-Meyer-Olkin for all the variables is 0.711 (acceptable for this test).

Figure 3 obtained through cluster analysis, considering the factor1 as variable, shows that several European Union regions considered in this study may be grouped into three main clusters. In this figure ‘G’ represents the group and ‘n’ the number of observations in each group.

The results presented in Table 7, obtained following Stata (2019) procedures for cluster analysis, for the three clusters already identified reveal that in cluster 1 it is possible to find Bulgaria, the Czech Republic, Lithuania, Hungary and some regions from France and from Poland. Poland appears alone in cluster 2, showing a context which is different to the other regions and countries.

In cluster 3 it is possible to find Cyprus, the regions from Bulgaria, the Czech Republic and Hungary and the remaining regions from France and Poland.

The DEA in this section was performed through DEAP (2019) software, considering a model based on the Cobb–Douglas developments, where the output is the standard output (euros) and the inputs are the utilized agricultural area (in hectares and as a proxy for the capital), the labour (annual work units) and the fresh surface water withdrawal (million cubic metres) for farming irrigation (to analyse the possibility of efficient savings for agricultural research). All these variables were considered disaggregated at national and regional level and only to 2010 (for better compatibility of several series). The selection of the countries and regions was also constrained by the availability of data in the mainstream databases.

The mathematical formulation to perform the DEA analysis based on the Cobb–Douglas model described before is well described, for example, in Coelli (1998).

This analysis will only apply to clusters 1 and 3 identified in the previous section, because for cluster 2 there was only Poland found as DMU. An analysis oriented input was considered, with variable returns to scale and DEA multi-stage.

Table 8 demonstrates that it is possible to improve the output of Bulgaria, the Czech Republic, Lithuania and of regions from France (Poitou-Charentes, Midi-Pyrénées, Rhône-Alpes). For example, in Rhône-Alpes it is possible to efficiently improve the standard output by about 70%. On the other hand, the French regions of Champagne-Ardenne, Bourgogne, Bretagne, Aquitaine and the Polish region of Lubelskie seem to be those which are more efficient in cluster 1. Specifically for the surface water withdrawal for agricultural irrigation, the situation is more critical for the French region; Centre, Pays de la Loire, Poitou-Charentes and Midi-Pyrénées where almost 100% of the water withdrawal could be saved (since the projected values are almost 0% of the original values). It is worth noting that the Bulgarian and Hungarian global contexts deserve special attention from the several stakeholders. In general, the frameworks for the utilized agricultural area and labour force follow a similar pattern and are not so problematic as those shown for fresh surface water. In any case, the Bulgarian, Czech and Hungarian contexts, for example, reveal that it is potentially possible to improve significantly the efficiency in the area and labour use.

For cluster 3, Table 9 shows that two regions from the Czech Republic (Praha and Moravskoslezsko), several regions from France, Cyprus and two Polish regions (Malopolskie and Slaskie) are the most efficient DMU. Concerning water withdrawal for agricultural irrigation the more dramatic cases are those from the Bulgarian regions (as it is for the other inputs, confirming the findings stressed for Table 8), some regions from Czech Republic (Strední Cechy and Severozápad, for example), some French regions (Auvergne, Languedoc-Roussillon, Provence-Alpes-Côte d'Azur and Corse) and the Hungarian regions of Észak-Alföld and Dél-Alföld. As referred to before, for cluster 1, in general, the contexts for the utilized agricultural area and labour force display a similar picture and are not so problematic as those presented for surface water; however, in some cases, the potentialities for savings in the labour force are more relevant, namely, in the Hungarian and Polish regions. This is also observable in Table 8 and in Figure 2.

Of course, despite cluster analysis being done to obtain homogeneous DMUs, following, for instance, Martinho (2020), there are still significant differences between the regions and countries considered in each cluster. In any case, the results presented and discussed here may provide an interesting base for the several stakeholders, namely, farmers and policymakers, in line with that argued, for example, by Salmoral et al. (2017). These findings may allow implementation of new and adjusted strategies to improve water withdrawal for agricultural irrigation, with relevant positive externalities for the environment and economic sustainability for farms.

Indeed, the amount of water withdrawal for agricultural irrigation and how effectively it is used, as shown in the statistics from the Eurostat, should be a motive for concern, as stressed by the Aquastat (2019). In general, the surface water and, in some cases, the labour force, are used less efficiently in the agricultural sector than, for example, the utilized agricultural area. The efficient use of water depends, in fact, on several factors, as stressed by Nazari et al. (2018).

The more dramatic cases are those from some French regions and from the Bulgarian and Hungarian contexts. The French context is paradoxical, because it has the most and less efficient regions, for surface water use for agricultural irrigation. Nonetheless, the French efficiency weaknesses are more focused on the surface water withdrawal and the Bulgarian and Hungarian fragilities are more scattered over the farm structures, involving the area and the labour management. This context highlights the diversity of realities inside the European Union, as shown by Martinho (2017).

The main objective of this study was to analyse the efficiency in water withdrawal for agricultural irrigation in the regions and countries of the European Union. For this, qualitative analysis of the literature was performed and statistical information from the Eurostat was considered over the period 2010–2012. These data were analysed through DEA, after factor and cluster analysis. The factor analysis allowed factors to be obtained without presenting problems of multicollinearity and the cluster analysis allowed for finding groups of more homogeneous DMUs (important aspects for the efficiency of the investigation). The variables considered were the standard output (as output), the utilized agricultural area, the labour and the water withdrawal. For water withdrawal, two series were considered, one for the fresh surface water and the other for fresh groundwater. Considering the difficulties in finding robust series for the water withdrawal, the factor, cluster and DEA analyses were performed only considering fresh surface water, for the year of 2010 and for the countries and regions with statistical information in the Eurostat. In fact, it will be important that public institutions improve the availability of statistical information concerning water withdrawal and its use by the agricultural sector.

The literature review, complemented with bibliometric analysis through the Atlas.ti software, highlights the following three main topics about efficient water management in the agricultural sector: irrigation efficiency, water shortage and water quantification. In fact, it is fundamental to improve irrigation efficiency, namely, with new approaches and technologies, where farmers have a determinant role. The questions regarding improvements in water efficiency gain more relevance when the water scarcity or shortage increases due to climate change and global warming. The new approaches for water quantification may also provide an important contribution.

It is important to highlight in this section that the data analysis shows that a relevant part of the water withdrawal for agricultural irrigation is from fresh surface water and that the factor and cluster analysis allowed identification of three clusters for the countries and regions considered in this study (the selection was constrained by the availability of statistical information in the mainstream databases).

Finally, the data envelopment analysis reveals that there are interesting cases of efficiency (including water withdrawal efficiency) that should be considered as references in a process of benchmarking that could be performed within the European Union, considering the specificities of the agricultural sector and the importance of an efficient management of water. On the other hand, there are some cases that deserve special attention, such as those, for example, from Bulgaria and some French regions, where, in some circumstances, almost 100% of the water withdrawal could be saved (of course, in an optimized perspective).

This study highlighted the main aspects related to efficient water management in the agricultural sector, making it possible to identify European countries and regions that may be considered as benchmarks for the less efficient regions and quantifying the potential possibilities of savings in the area, labour and farm water use. The main limitation of this work is related to the availability of statistical information in the mainstream databases. In future research, the approaches presented in this study may be applied to other parts of the world outside the European Union, by benchmarking the results obtained with those demonstrated here.

This work is financed by national funds through FCT – Fundação para a Ciência e Tecnologia, I.P., under the project UID/Multi/04016/2019. Furthermore we would like to thank the Instituto Politécnico de Viseu and CI&DETS for their support. This work is supported by national funds, through the FCT – Portuguese Foundation for Science and Technology under the project UID/SOC/04011/2019.