USING WATER FOOTPRINT CONCEPTS FOR WATER SECURITY ASSESSMENT OF A BASIN UNDER ANTHROPOGENIC PRESSURES

The evaluation of water shortages and pollution levels is crucial for watershed management and sustainable development. This paper proposes a water footprint (WF) sustainability assessment approach to analyse water security in a river basin under human pressures. The methodology involves a comprehensive assessment of the current water security at different spatial and temporal levels, and identifies suitable response formulations to achieve sustainability. Field surveys and measurements (streamflow, water quality) were carried out, and the Soil and Water Assessment Tool model was used for assessing water balance components and water quality. The study was carried out in the Canale d’Aiedda river basin (Taranto, Italy), which is part of the ‘area of environmental crisis’ of Taranto, which requires remediation of surface water, groundwater, soil and subsoil. Considering all the anthropogenic activities in the basin, including agriculture and the treated effluent disposed of via wastewater treatment plants (WWTPs), the average WF was 213.9 Mm 3 y -1 , of which 37.2%, 9.2% and 53.6% comprised respectively for WF green , WF blue and WF grey . The WF sustainability assessment revealed that pollution was the main factor affecting surface water security. In particular, point sources contributed with 90% to the total WF grey , and lower pollutant thresholds should be fixed for effluent from WWTPs in order to increase water quality of the receiving water body. In addition, for assuring water security the extension of the natural areas should be increased to support biodiversity in the river basin and soil management strategies should be improved to allow more water to be retained in the soil and to reduce nutrients in runoff. This study demonstrates that the WF sustainability assessment is a feasible approach for integrated water resources management, as well as offering a much broader perspective on how water security can be achieved in a Mediterranean basin affected by multiple anthropogenic stressors.


Introduction
Fresh water is essential for human life, for a thriving economy and for supporting natural ecosystems (Oki and Kanae, 2006). Climate change, increasing demand, and pollution could put the availability of water resources for future generations at risk (De Girolamo et al., 2017a;Pophare et al., 2014;Rosa et al., 2018). International organisations, aware of this risk, have introduced the concept of water security (GWP, 2019;UNESCO and UNESCO i-WSSM, 2019). This has received increasing attention in policy and scientific debates in recent decades (Cook and Bakker, 2012). Water security is defined as "the availability of an acceptable quantity and quality of water for health, livelihoods, ecosystems and production, coupled with an acceptable level of water-related risks to people, environments and economies" (Grey and Sadoff, 2007). The term refers to economic welfare, social equity, long-term sustainability and water-related risks (Hoekstra et al., 2018). Water security is a conceptual framework for analysing water systems that is able to integrate multidisciplinary concerns across spatial scales (from local to global). The main challenges of this broad conceptual approach are the implementation and management processes (Cook and Bakker, 2012).
Water security assessments based on the concept of the water footprint (WF) are becoming globally largely accepted (Giri et al., 2018;Hoekstra, 2017;Veettil and Mishra, 2016). The WF quantifies the direct and indirect volume of water required by human activities in a specific geographical area during a unit of time (Hoekstra et al., 2011). It is partitioned into three components: green, blue and grey. The green WF (WFgreen) refers to the amount of precipitation temporarily accumulated on vegetation and/or in soil, and that is available for evapotranspiration through the vegetation and soil (Falkenmark, 2003). The blue WF (WFblue) is the volume of fresh water provided along a supply chain of products and/or services (Veettil and Mishra, 2016). The grey WF (WFgrey) refers to the degree of water pollution induced by human activities, and is defined as the volume of fresh water required to dilute a load of pollutants so that the receiving water body can assimilate those loads without compromising the achievement or maintenance of specific water-quality standards (Franke et al., 2013).
In recent decades, the WF concept has been used to evaluate the sustainability of water resources management (Mikosch et al., 2020), and the common spatial unit considered in such evaluations is the river basin (Liu et al., 2017;Zang et al., 2012). The WF sustainability assessment is a reliable way to estimate current water shortages and pollution apportionment, from point sources (PSs) and diffuse sources (DSs), and for identifying local hotspots in a river basin Fig. 1: Location, land use and hydrological soil groups identified in the study area. The thick black line delimits the area contributing to the surface runoff, which was divided into 40 subbasins (thin black line) for the purpose of applying the SWAT model. Regional Park and Forest Reserve are identified (thin red line).
In the basin, the main economic activity is extensive agriculture. A limited number of breeding farms is mostly localised in the north of the study area. Vineyards (36.3%), olive groves (24.5%) and durum wheat in rotation with herbage and fallow land (28.1%) constitute the main land uses (Fig. 1). Citrus groves and vegetables represent minority land uses. These latter, in addition to the vineyards and olive groves, are irrigated. The main irrigation source is deep groundwater. The urban area accounts for 3.9% of the basin. Deciduous (e.g. holm oak) and coniferous forest are present (2.7%), as well as pasture (e.g. Mediterranean maquis) (2.4%) and bushes and shrubs (e.g. myrtus, mastic bush) (0.9%). These natural areas mainly fall within Regional Park and Forest Reserve (i.e. Terra delle Gravine, Bosco delle Pianelle).
The river network is heavily modified, with concrete river banks being present (D'Ambrosio et al., 2019). In addition, the hydrological regime has been altered by the discharge from three WWTPs -Montemesola (W1), Monteiasi (W2) and San Giorgio Ionico (W3)the contributions from which assure permanent downstream streamflow (Fig. 1).
Meanwhile, an intermittent regime is recognisable upstream.

Material and methods
The integrated modelling framework proposed for water security assessment consists of three steps: 1) evaluating the water balance components, water quality and source apportionment through modelling; 2) assessing the WF sustainability; and 3) formulating responses to achieve water security in hotspot areas. The second and third steps are based on the WF concepts of green water scarcity (WSgreen), blue water scarcity (WSblue) and water pollution level (WPL) (Hoekstra et al., 2011), which were analysed at the subbasin and basin scales, considering monthly and yearly time scales. Meanwhile, the hydrological and water-quality model was used to evaluate the variables (i.e. water consumption, nutrient load adducted to the river for each crop) necessary for the WSgreen, WSblue and WPL calculations ( Fig. 2). Each step is explained below. The procedure was applied to the Canale d'Aiedda Basin.
The input data required by the SWAT, and used in the current study for setting up the model, are provided in Table 1.
Details on discharge volumes and the main pollutant concentrations are provided in Table 2 for each PS (i.e. treated effluent from WWTPs, springs).  the limited available data could not be meaningfully split into two subsets to be used for the calibration and validation, respectively. Indeed, Arsenault et al. (2018) pointed out the need to use the full time-series in the calibration step and to disregard the validation aspects under certain conditions. In this case, to make a robust calibration, a split-in-space strategy was adopted by using measured data in two gauging stations (A, and B in Fig. 1). Details on the monitoring activities (from which the data for the model calibration were obtained) are reported in D' Ambrosio et al. (2019Ambrosio et al. ( , 2020. Table 3 provides the mean results from gauges A and B. The Sequential Uncertainty Fitting (SUFI-2) procedure in SWAT-CUP (Rouholahnejad et al., 2012) was used to autocalibrate 104 model parameters for both gauges (D'Ambrosio et al., 2020). The calibration protocol indicated by Abbaspour (2015) was applied. The model performance was evaluated using the coefficient of determination (R 2 ), the Nash-Sutcliffe efficiency (NSE) coefficient and the percent deviation (PBIAS), with the criteria of acceptability being fixed as suggested by Moriasi et al. (2007).
Following the calibration, the SWAT was run without crop irrigation and WWTP input, in order to calculate the WSgreen and WSblue, respectively (Fig. 2). Other simulations were executed that considered different WWTP effluent conditions in order to identify a response formulation towards achieving sustainability, where necessary.

Green water scarcity
Although Savenije (2000) has already underscored the importance of the WFgreen sustainability assessment, most of the studies on water scarcity have focused only on blue water resources . According to Hoekstra et al. (2011), the WFgreen sustainability assessment is based on the evaluation of WSgreen, which is defined as the ratio of the total WFgreen associated with crop production to the green water availability (WAgreen) within a certain period. In the study area, vineyard, olive grove, citrus groves, durum wheat and herbage were the productive crops. The total WFgreen represents the volume of green water consumed through evapotranspiration processes during the growing period by the productive crops cultivated in the analysed spatial unit (Pellicer-Martínez and Martínez-Paz, 2016). It is equal to the actual crop evapotranspiration of productive crops, assuming that the soil does not receive any irrigation (ETa no irr ) (Hoekstra et al., 2011). To estimate the ETa no irr , the SWAT model was run for the current land use using the best model parameters' values identified with the calibration procedure ("calibrated SWAT model") without crop irrigation.
In the study area, durum wheat and herbage are the only temporary crops. The former is sown in late autumn (1 st December in the SWAT simulation) and harvested in early summer (30 th June in the SWAT simulation), the latter sown in autumn (1 st November) and also harvested in early summer (30 th June). Thus, only those months (from December to June and from November to June) were considered in the WFgreen assessment of durum wheat and herbage, respectively.
In the remaining months, the soil was bare. Meanwhile, vineyard (harvesting date: 15 th August), olive grove (harvesting date: 15 th October), and citrus groves (harvesting date: 15 th October), contributed throughout the whole year in determining the WFgreen, being permanent trees.
The WAgreen was calculated by: ETgreen is the volume of green water available at the basin or subbasin scale (actual evapotranspiration for the current land use excluding irrigation). Similarly to the flow requirements to support aquatic ecosystems, the ETenv is defined as the green water flow to be reserved for nature, needed to support biodiversity. In the present work, land reserved for nature was very limited (6%), thus, it was assumed to achieve the Aichi Biodiversity Target (ABT) 11, which entails expanding the protected area network to at least 17% of the terrestrial world by 2020 (www.cbd.int/sp/targets).
ETenv was computed as the actual evapotranspiration of natural vegetation present in the area (i.e. bushes and shrubs, forest and pasture) that was incremented in order to achieve the ABT 11 target (17% of the catchment area). In particular, it was supposed to convert part of arable lands (i.e. durum wheat and herbage) falling within Regional Park and Reserve ( Fig. 1) into forests. The computation of ETenv was performed by using the SWAT model assuming the default values of crop coefficient for bushes and shrubs, forest and pasture (garrigue). Meanwhile, ETunprod is the evapotranspiration from unproductive land (i.e. urban areas, reservoirs), which is not available for cultivation, and it also includes fallow land that is essential for restoring and improving soil fertility for next cultivation period.  evapotranspiration from unproductive land (ETunprod) and actual evapotranspiration from current natural vegetation and additional area to be reserved for Nature to achieve the ABT 11 target (ETenv). The size of the rectangles is linked to the contribution to determine the ETgreen of each land use, of which the percentage is provided.
Being a ratio, the WSgreen indicates the degree of green water use throughout the analysed spatial unit. A WSgreen of >1 means that the available green water is fully consumed by productive crops (Hoekstra et al., 2011), which should be partially converted in natural areas to support biodiversity; < 1 (or =1) indicates a suitable WSgreen, where there is enough green water to meet the productive and natural crop demands.

Blue water scarcity
The environmental sustainability of the WFblue is evaluated by means of the WSblue., which is the ratio of the total WFblue to the blue water availability (WAblue) (Hoekstra et al., 2011). Based on data availability, this study focused only on the surface water, ignoring the sustainability of groundwater use. In addition, only the agricultural water use was considered as water for industrial, commercial and domestic uses was imported from outside the Canale D'Aiedda Basin.
The total WFblue is the amount of fresh water consumed in crop production (i.e. irrigation). This fresh water is 'consumed' because it does not return to the source in the form of return flow, being used up via evapotranspiration, incorporated into products and/or returned to another catchment or to the sea. In the study area, only the olive groves, vineyards and citrus groves are irrigated. The WFblue is obtained from the difference between the actual crop evapotranspiration, considering current management practices (ETa) and ETa no irr (Hoekstra et al., 2011). Thus, the calibrated SWAT model output was used to obtain the ETa.
The WAblue is the maximum volume of water that can be abstracted from a river without harming its ecosystems (Bonamente et al., 2017). This was assessed through the following equation (Hoekstra et al., 2011): Ract is the actual runoff from the basin or subbasin, which was obtained from the calibrated SWAT model output (including WWTP discharge). EFR is the environmental flow requirement, needed to preserve the ecological integrity of downstream aquatic ecosystems. Depending on data availability (Tharme, 2003), this can be assessed by a multitude of methods that can be divided into four categories (i.e. hydrological, hydraulic, habitat simulation and holistic). The 'presumptive environmental flow standard' proposed by Richter et al. (2012), and earlier applied in many WF sustainability assessment studies (e.g. de Miguel et al., 2015;Kaur et al., 2019;Veettil and Mishra, 2018), was used here. According to this standard, the EFR accounts for an 80% share of the natural flow regime (Rnat). Thus, the calibrated SWAT model was run without considering WWTP discharge, in order to obtain the Rnat and then the EFR (=0.8 Rnat).
The WS blue nat was assessed without including PS discharge, by substituting Rnat with Ract in Eq.
(2) to give the WA blue nat .

Water pollution level
The WPL is the fraction of the waste assimilation capacity consumed at the analysed spatial scale (Hoekstra et al., 2011). It is quantified by dividing the total WFgrey by the grey water availability (WAgrey). The WFgrey refers to the amount of fresh water needed to dilute the pollutant load discharged into a receiving water body such that the quality of the water body remained within the range of a fixed water-quality standard (Franke et al., 2013). Meanwhile, the WAgrey is the amount of fresh water that could be polluted without affecting the ecosystems of the receiving water body (Bonamente et al., 2017). Due to a lack of data, this study focused on the surface water, ignoring any pollution in the groundwater.
In the study area, the surface water quality is impaired by both agriculture (i.e. DS) and WWTP discharge (i.e. PS) (D'Ambrosio et al., 2019). Thus, the total WFgrey was calculated as the sum of the WFgrey associated with the DSs (WF grey DS ) and PSs (WF grey PS ).
WF grey DS (m 3 ) was calculated thus (Hoekstra et al., 2011): , where L DS (kg) is the diffuse pollution load adducted to the river network, and Cmax and Cnat (kg m -3 ) are the maximum allowed and natural concentrations of the pollutant in the water body, respectively.
In Eq. has provided five different thresholds (from poor to high water quality) for N-NH4, N-NO3 and TP, as supporting chemical parameters for ecological status evaluation. In this study, the 'good' TP environmental water-quality standard was adopted (Cmax TP =0.0001 kg m -1 ). Meanwhile, the standard proposed by Liu et al. (2017) was used for TN (Cmax TN =0.003 kg m -3 ), since the national legislation has not fixed a threshold for that.
Several values of Cnat have been reported in the literature for both TN (Cnat TN ) and TP (Cnat TP ) due to huge differences in their environmental characteristics (Hejzlar et al., 2009;Meybeck, 1982;Smith et al., 2003). In this study, since it was not possible to measure Cnat because of the human disturbance, Cnat TN and Cnat TP were assumed to equal 0.00065 kg m -3 and 0.000013 kg m -3 , respectively (D'Ambrosio et al., 2020).
WF grey DS was calculated separately for TN (WF grey DS,TN ) and TP (WF grey DS,TP ), and the highest value was assumed to be the final value.
WF grey PS (m 3 ) was assessed using: , where L PS (kg) is the pollution load resulting from the WWTP effluent discharged within the river network, which was provided by the calibrated SWAT model, considering the measured concentrations of effluent (Ceffl TN , Ceffl TP ) and the regulatory limits fixed by national legislation (Ceffl TN* , Ceffl TP* ) ( Table 2). The L PS was obtained by subtracting the L DS from the total pollution load (L, L TN , L TP ) provided by the model, which was run considering both the DS and PS. Thus, L PS,TN , L PS,TP , L PS,TN* and L PS,TP* were considered in Eq. (4).

Hydrological and water-quality model results
The model performances ranged from acceptable to very good (Moriasi et al., 2007) (Table 4)  The specific crop contributions were analysed in Section 4.2. Potential evapotranspiration was about 1200 mm y -1 .
Meanwhile, the Rnat, on average, accounted for a 12.0% share of the annual rainfall. Its annual volume varied between 3.7 Mm 3 (2017) and 24.4 Mm 3 (2015), with a mean value of 17.2 Mm 3 . The mean annual contribution of the WWTP effluent disposed of in the river network was estimated to be 3.7 Mm 3 , which resulted in an average increase of 21.7% in the streamflow at the basin outlet. In particular, the actual streamflow (Ract) ranged from 6.6 Mm 3 (2017) to 28.2 Mm 3 (2015) (mean 20.9 Mm 3 ).
The main components of the annual average water balance simulated by the SWAT model for the current land use are shown in Fig. 4. In the graph, the components are allocated among the users. Ract is differentiated in EFR, WWTPs contribution and amount available for human uses (Rnat-EFR), meanwhile the deep aquifer recharge is differentiated in water abstraction for irrigation and net recharge. Total actual evapotranspiration is differentiated for productive land (including irrigated crops), unproductive land (non-utilizable and fallow land) and natural areas. to 1.65 kg ha -1 y -1 ) and 0.04 kg ha -1 y -1 (from 0.02 to 0.10 kg ha -1 y -1 ), respectively. The higher specific values were associated with bushes and shrubs, with a mean value of 4.1 kg ha -1 y -1 for TN and 0.1 kg ha -1 y -1 for TP. This was due to their proximity to the river network and to grazing activities.

Water footprint sustainability assessment results
The annual total water footprint of the Canale d'Aiedda Basin, which was obtained by summing the total WFgreen, WFblue and WFgrey, ranged between 120.6 and 259.  The mean monthly variations in the WF components and WF sustainability assessment indicators are shown in Fig.6. The mean monthly WFgreen ranged between 2.5 Mm 3 (July) and 14.3 Mm 3 (April), depending on climate and vegetation cover (Fig.6a). In particular, temporary crops (i.e. durum wheat and herbage) from July to October did not contribute to the WFgreen, since that ground was fallow in those months. Also, the ETgreen and WAgreen showed a marked seasonality, with higher values being recorded in spring. Meanwhile, the monthly variability of the ETenv was low, being between 0.9 Mm 3 (December) and 2.9 Mm 3 (April).
At the basin scale, the mean (2011-2017) annual WSgreen was 1.2. However, when the temporal variability was assessed at the monthly scale, its values fluctuated between 1.12 (May) and 1.45 (August).
With respect to the mean monthly WFblue, the results were >0 from May to September, when the olive groves and vineyards received irrigation. The lowest value was registered in September (880.4 m 3 ), the highest in July (6 Mm 3 ). The Ract, Rnat and EFR varied little over the months. Specifically, on a monthly time scale, they averaged between 0.9 and 3.0 Mm 3 , 0.6 and 2.6 Mm 3 , and 0.5 and 2.1 Mm 3 , respectively, following the precipitation regime.
Conversely, the WSblue and WS blue nat were characterised by a relative discrepancy between the summer months and the rest of the year. Indeed, the WSblue varied between 0 and 13.19 (August) at the basin scale and at a monthly temporal scale. The maximum value increased to 100.67 for the WS blue nat . In August, throughout the studied period, the WSblue varied from 6.88 (2012) to 27.65 (2017), whilst the WS blue nat ranged from 16.48 to 518.10, respectively. In terms of the large interannual variability, the standard deviation associated with August was the highest for both indicators (Fig.6b).
The mean (2011-2017) annual WSblue and WS blue nat were found to be 3.0 and 8.46, respectively. Therefore, if irrigation were provided from surface water, the Canale d'Aiedda Basin would experience severe blue water scarcity. However, deep groundwater constituted the main source for irrigation in the study area.
The mean monthly WFgrey varied from 6.8 Mm 3 (October) to 14.3 Mm 3 (March) (Fig.6c). Specifically, the highest value was registered when peak pollutant concentrations in discharged effluent overlapped with the runoff from fertiliser applied in February to the durum wheat and vineyards. Meanwhile, the WF grey DS and WF grey PS ranged between 0.03 Mm 3 (August) and 4.6 Mm 3 (February), and between 6.0 Mm 3 (September) and 12.0 Mm 3 (March), respectively. From April to August, more than 99% of the WFgrey was due to WF grey PS . If the limits fixed by national legislation for WWTP treated effluent (i.e. Ceffl TN* , Ceffl TP* ) were adhered to, the WF grey PS * would remain at the same level (4.5-6.7 Mm 3 ) for almost the whole year.
With respect to the WPL, the agricultural activities were sustainable at the yearly average, and also when the subbasin scale was considered. Only Subbasin 9 had a mean annual value of WPL DS >1 (Fig.7c, Table 5). The WWTP effluent consumed the capacity to assimilate the pollutant load in downstream subbasins (i.e. Subbasins 14,19,23,24,27,29,30,31,34 and 35 in Fig. 1), considering both Ceffl TN and Ceffl TP (WPLin Fig.7d) and Ceffl TN* and Ceffl TP* (WPL* in Fig.7e). Depending on the month, TN or TP were alternatively found to be the critical pollutant for both the DS and PS.
In order to obtain a sustainable pollution level (WPL S in Fig.7f), the discharge from W1, W2 and W3 would have to respect the desirable TN and TP concentrations (Ceffl TN_S , Ceffl TP_S ) obtained from the iterative procedure described in Section 3.2.3. These concentrations are reported in Table 6.   Table 6: Desirable TN (Ceffl TN_S ) and TP (Ceffl TP_S ) concentrations in WWTP effluent (W1, W2, W3) required to obtain a sustainable for irrigation purposes could be promoted in a circular-economy context, since this would also promote a reduction in fertiliser usage, as well as the partial satisfaction of irrigation demand (De Girolamo and Lo Porto, 2020).
The suitable thresholds (Ceffl TN_S , Ceffl TP_S ) identified in the current study improve on those obtained by a previous study in the same study area (D'Ambrosio et al., 2020). Indeed, D'Ambrosio et al. (2020) identified desirable concentrations of TN and TP equal to 8.1 and 0.3 mg L -1 , respectively, that were evaluated at the basin scale.
Concerning agricultural activities at the basin scale, February, September, November and December were found to be critical months, with the WPL DS being 2.30, 1.18, 1.36 and 1.16, respectively. In those months, fertilisation was carried out on the vineyards (February, November), durum wheat (February, December) and herbage (April, September) throughout the basin. If the yearly average of the WPL DS at the subbasin scale is considered, low unsustainability (WPL DS =1.1) was associated with only Subbasin 9 (Fig. 1,7c). This is due to the contribution of a plurality of factors, such as the high runoff potential of the soil, land use (29.2% olive, 16.5% vineyard, 10.0% herbage) and slope (about 3.7°). Hence, fertiliser strategies aimed at minimising nutrient runoff (i.e. precision agriculture) should be adopted, especially in hotspot subbasins and in critical months (D'Ambrosio et al., 2018a;De Girolamo et al., 2017c). Replacing part of the chemical fertiliser with mycorrhizal fungi  or with treated, reusable wastewater (De Girolamo and Lo Porto, 2020) could be effective ways to reduce the WF grey DS . The involvement of local farmers and dealers, as well as a detailed economic analyses, are fundamental in supporting any decisions about the agricultural practices to be promoted and the pursuit of win-win solutions (Lovarelli et al., 2016;Mekonnen and Hoekstra, 2010).
Water security assessments should not neglect green water resources since these are critical and limited; as with, blue water, they should explicitly be taken into account because both are related to the amount of precipitation, which is limited in time and space Schyns et al., 2019). The limited use of the WSgreen indicator in watershed management is due to the lack of a standardised methodology for computing the WAgreen (D' Ambrosio et al., 2020;Quinteiro et al., 2018). Different formulas have been proposed and used for its assessment in the few studies that have evaluated it (Giri et al., 2018;Kaur et al., 2019;Pellicer-Martínez and Martínez-Paz, 2016;Salmoral et al., 2017;Schyns et al., 2019;Veettil and Mishra, 2018), which has resulted in comparisons becoming meaningless.
The conceptual approach of the WSgreen developed by Schyns et al. (2019) and adopted in the present work refers to the competition for green water resources, which should support both the natural ecosystem and crop production. The WSgreen indicator quantifies the degree of human appropriation of the green water flow. In their study, the authors quantified the maximum sustainable WAgreen considering agroecological suitability and accessibility of land, biophysical constraints to intensifying land use, and biodiversity conservation needs (ABT 11). In that approach, increasing WSgreen means that reduced green water resources remain for nature. At the basin scale, if WSgreen is >1 means that the target fixed for supporting the biodiversity was not achieved and the WSgreen is defined "unsustainable". Mishra (2016, 2018) and Giri et al. (2018) adopted a different approach for estimating the WAgreen, which was the amount of initial soil water content and it is computed as the difference between soil moisture at the root zone and wilting point. In this approach, there is no target to be achieved for biodiversity.
The final aim of the study should guide in choosing the most suitable methodology. Thus, the second approach (Veettil and Mishra, 2016) could be more suitable in studies aimed at assessing the impact of land-use or climate change on the WAgreen at a local scale. Meanwhile, the studies evaluating water security using the concept of the WF should privilege the approach developed by Schyns et al. (2019). However, an integration of the two approaches is desirable.
At the basin scale, the mean annual WSgreen is 1.20. This value indicates that reduced green water resources remain for nature. The value is in line with the overshoot of the WAgreen identified for the study area by Schyns et al. (2019) and is higher than the value estimated at the national level (0.9-1.0). At the subbasin scale, some subbasins face water insecurity where WSgreen is >1. These subbasins have tree crops (i.e. olive groves and/or vineyards) as the main land cover, and all the WAgreen has been allocated to human (agricultural) use. Meanwhile in the subbasins with the presence of natural areas WFgreen equals WAgreen. The Canale D'Aiedda basin has no potential remaining to increase rain-fed agricultural production. In order to locally increase the WAgreen, a conversion of part of the agricultural land in natural areas is proposed in this study in order to enlarge the existing protected areas. However, additional investigations are needed to explore in detail this possibility. Indeed, protected areas should be integrated into a wider region considering ecological connectivity and networks and should be established and managed in collaboration with local communities, with regional and national authorities. Soil management practices should be improved to enable more water to be retained in the soil, which can then be taken up by crops and recycled via transpiration (Schyns et al., 2019). The best management practices would also reduce the WFblue because less water would be necessary for irrigation to prevent the plants from experiencing water stress. In particular, no-and reduced-tillage systems (which increase carbon sequestration in soil and improve soil organic matter), or the application of mulch, could limit soil evaporation and increase transpiration efficiency, thereby contributing to biomass growth. However, an optimal irrigation scheduling and new efficient technologies and equipment should be adopted for improving water use.
The WSblue exhibited different trends, which were mainly determined by the seasonality of irrigation water consumption. Indeed, in the study area, irrigation was predominantly provided to the vineyards and olive groves from May to September. Thus, at the basin scale, that indicator resulted in values >0 only in those months, increasing progressively from 3.73 to 13.19. These values ranged from 10.90 to 100.67 when the WWTP effluent contribution was neglected (WS blue nat ). Therefore, surface water (WAblue, WA blue nat ) was not sufficient to satisfy the irrigation requirements of the study area, and using deep groundwater remained necessary because there is no irrigation board in this area.
However, surface water could be used for reducing the quantitative pressures on groundwater resources (Casella et al., 2019). Specifically, in some subbasins, the annual irrigation water requirement could be completely satisfied by both WAblue and WA blue nat , even when taking their monthly variability into account. In these subbasins, water conservancy infrastructure could be used to accumulate surface water during flood events (thereby also reducing the flood risk) to be released in dry periods to fulfil irrigation-water requirements (Cai et al., 2020).
In summary, the implementation of strategies aimed at reducing the three WF components are fundamental to increasing water security in the Canale d'Aiedda Basin. Given that the WFgrey constitutes more than 50% of the total WF, it is clear that improving wastewater treatment and management, and reducing nutrient runoff from agricultural land, is of paramount importance.

Advantages related to the assessment of water security using water footprint indicators and hydrological and water-quality models
Water security is broadly recognised as an important and urgent policy challenge, which can be achieved by balancing human needs while safeguarding ecosystems and biodiversity (Bakker, 2012). In order to address this challenge, it is fundamentally important to quantify water security (Hoekstra et al., 2018). In this sense, indicators can be a powerful tool. They can help to define the current situation as well as changes over time, prioritise actions and investments and measure their effectiveness, and stimulate policy-makers because they can be shown a direct link between actions and results (Jensen and Wu, 2018). Indicators are considered meaningful and feasible for use in policy-making only if they are credible, valid and salient (Hoekstra et al., 2018). In other words, they need to be scientifically valid and technically robust to be legitimised and accepted by stakeholders and recognised by the policy-makers who should be aware of them. In addition, the assessment of such indicators should be based on data that are available from trusted sources and that can be collected within time and budgetary constraints (Jensen and Wu, 2018).
WF indicators are characterised by the features mentioned above and have become the main international reference for evaluating the sustainability of water use since the ISO 14046 norm was adopted (Pellicer-Martínez and Martínez-Paz, 2016). In addition, WF concepts can be used for understanding the complexity and temporal-spatial variability of water security through the adoption of a system-dynamic perspective based on the pressure-state-impact-response schematisation of social-environmental systems (Hoekstra et al., 2018). They allow the evaluation of water supplies and water-use efficiency, as well as aspects of quality. Thus, the traditional idea of water security assurance, which has historically focused on water-quantity items, ignoring the fact that pollution can compromise water use (David da Silva et al., 2020), can be improved through use of the WF sustainability concept . To guarantee water security, collective actions should be taken to effectively reduce the WF of human activities, moving them towards sustainable levels . Interventions via different actors along the supply chains, and at different spatial and temporal scales, are essential, and the use of modelling approaches at the watershed scale can enable waterresource managers to address the challenge of water security by enabling them to foresee the effects of particular response formulations that have been identified as being able to reduce the multiple pressures that may currently affect the analysed water system (Giri et al., 2018).
Hydrological and water-quality models (such as SWAT) are robust tools for WF assessment, since processing the model outputs makes it possible to calculate all the WF components, together with the maximum sustainable values of the WF at the different spatial and time scales. This allows a WF sustainability analysis to be performed by direct comparison (Pellicer-Martínez and Martínez-Paz, 2016). Models can be used to address socio-environmental issues relating to actual land-use management, and/or for predicting the effects on water security of alternative scenarios and the implementation of best management practices (Giri et al., 2018;Malagó et al., 2017;Pulighe et al., 2019). This issue is particularly pressing for Mediterranean basins, considering that climate change is expected to have a great impact on the water balance, with an increase in temperature, a reduction in rainfall and changes in river hydrological regimes, all of which will likely negatively impact water security (D'Ambrosio et al., 2018a;De Girolamo et al., 2017a).
For all these reasons, the integrated modelling framework proposed in this study could have great utility in integrated water-resources management, since it allows a better understanding of how (hydrological and water-quality processes), where (environmental hotspots) and when (time) water scarcity may occur across a basin, considering both the current condition and different future scenarios and WF mitigation strategies (Giri et al., 2018).

Limitations of the study and future improvements
The concept of water security embraces multidisciplinary concerns related to economic welfare, social equity, sustainability and risk (Hoekstra et al., 2018), and thus presents a way of thinking about all water-related issues, such as safe drinking water, water scarcity, water pollution and flooding, in the context of an integrated framework (Sun et al., 2016). To capture the different facets of water security, a very large number of indicators have been developed. Jensen and Wu (2018) grouped these indicators into four main classes, related to resource availability (e.g. WF indicators, water storage capacity), access (e.g. water poverty index, drinking-water safety, water tariffs, water treatment capacity and coverage), risk (e.g. flood frequency and damage indicators) and institutional management capacity appraisal (e.g.
strategic planning, disaster management regulation). Thus, water security could be achieved by increasing resource availability and access, minimising water-related risks and enhancing management capacity. In addition, major pressures affecting water security can be divided into environmental and socioeconomic categories (Cai et al., 2020; a model could help to mimic the complexity of land-use systems in future applications. Other limitations affecting the WPL assessment resulted from only considering TN and TP as pollutants. The combined effect of coexistent pollutants, such as pesticides (Gil et al., 2017) and human and veterinary pharmaceuticals (Wöhler et al., 2020), should be analysed so as to better evaluate the actual WFgrey.
Similarly, the WPL assessment was limited to agricultural activities and WWTP effluent, which are the main activities in the basin; however, as a general procedure, all the anthropogenic activities carried out across the whole basin, such as by the industrial and domestic sectors (Liu et al., 2017), should be considered in order to enable an evaluation of the actual environmental sustainability level of the study area.
In addition, it is necessary to highlight that the WF sustainability assessment procedures needed to be improved in several aspects. Although the WF concept can be traced back to Hoekstra and Hung (2002), uncertainties remain concerning the standardisation of a single procedure to produce a consistent sustainability assessment. Indeed, the literature provides various methodologies for assessing WAgreen, EFR and WFgrey, which makes it difficult to compare the results of different studies. Schyns et al. (2019) recommended that future work focus on better estimating WAgreen, especially with regard to ETenv and ETunprod. Meanwhile, in the current study, the procedure proposed by Richter et al. (2012) for EFR accounting was adopted. However, a large number of methods exist for estimating EFR, and these can give significantly different results (Arthington, 2012). Finally, for the WFgrey assessment, the assumptions related to Cmax and Cnat significantly influenced the results and needed to be standardised (D'Ambrosio et al., 2018b;Liu et al., 2017). Further research on these issues is needed in order to standardise the methodology for WF sustainability assessments.
Finally and importantly, one of the major challenges is the translation of research results into coherent mechanisms of governance and ways of intervention, along supply chains and at different levels . Constructive synergies between researchers, decision-makers and practitioners are fundamental.

Conclusions
The concept of water security throughout a holistic approach aims to achieve sustainable development and human wellbeing. From a theoretical standpoint, water security has been well defined; however, an operational strategy is still in a critical phase, and several difficulties exist in integrating the concept into different levels of governance (at global, national and local scales). Indeed, the achievement of water security needs not only coordinated policies and good governance, but also interdisciplinary approaches, and models able to integrate its multi-sectoral aspects.
This study has provided a contribution that helps in the development of a procedure for implementing the water security concept at the local scale through a case study. The proposed approach was intended to be an operational tool for high spatial and temporal resolutions. It was defined by integrating the WF sustainability assessment methodology proposed by the Water Footprint Network with a hydrological and water-quality model and field surveys. Coupling modelling activities with the WF sustainability assessment procedure has proven effective in responding to water-security challenges because both quantitative and qualitative aspects can be evaluated.
The outcomes of this work have revealed critical issues in water security at the basin scale. While this spatial scale is appropriate as the basic unit for water resources management, this study has shown that a subbasin scale is needed to identify where and when management options have to be implemented in order to deliver water security. In the Canale D'Aiedda Basin--a basin representative of the Mediterranean climate--pollution is the main factor affecting water security, and so lower pollutant thresholds for WWTP effluent were identified as a means of increasing water quality. In addition, the natural areas should be increased to locally increase the WAgreen as well as agricultural management practices should be improved to enable the retention of more water in the soil and reduce nutrient runoff. Currently, surface-water availability is not sufficient to satisfy the irrigation requirements across the study area, and extracting water from deep groundwater remains necessary because there is no irrigation board operating in this area. This study also highlighted that the WF sustainability assessment procedure needed to be improved in several aspects.
The integrated framework developed in this study may have great utility in integrated water-resources management because it could support the implementation of a suitable water policy that would ensure water security in basins under