Coupling the water footprint accounting of crops and in-stream monitoring activities at the catchment scale.

In this work, a simple approach for calibrating the water footprint (WF) accounting of crops with in-stream measurements at the catchment scale was developed. The green and blue components of the WF were evaluated by performing a soil-water balance at a 10-day time-interval. The surface runoff was calibrated based on continuous streamflow measurements. Meanwhile, the grey component of the WF related to nitrogen use was quantified by means of the results from the in-stream monitoring activities. The methodology can be applied to any catchment where soil, land use, weather, agricultural practices, nitrogen balance and stream data are available. This methodological approach can support local authorities in the decision-making process for effective agricultural policy setting and water planning. •The WF accounting for an agricultural catchment is coupled with surface-water monitoring results•The green and blue WF are assessed by performing a soil-water balance•Surface runoff and grey water accounts are based on in-stream monitoring activities.

Specifications Table

Method details

The water footprint

The water footprint (WF) is a relatively new indicator, introduced by Hoekstra and Hung [2] to enable the quantification of water consumption and pollution, and to foster the implementation of more sustainable water-use practices. Galli et al. [3] included the WF in their ‘footprint family’, together with ecological and carbon footprints, as a suite of indicators useful in tracking human pressures on the planet from different aspects. The WF is a multidimensional indicator, which accounts for both the direct and indirect appropriation of freshwater resources.

WF assessments should be conducted at the river basin scale in the context of integrated water resource management aimed at sustainable development [[4], [5], [6]]. Despite this observation, agricultural WF accountings at the river basin scale are rare, due to the lack of reliable data at this scale, especially for arid and semi-arid regions, because of the high fragmentation of land use and variable adopted management practices [7].

Method objectives

This study aimed to give general guidance on how to perform a WF accounting of crop production at the catchment scale, in order to evaluate the sustainability of agricultural activities by considering both water quantity and quality. In particular, the defined methodologies aimed to: i) gather reliable data on land use and agricultural practices adopted within the catchment; ii) suggest a detailed program of surface-water monitoring, in order to better understand the hydrological and water-quality processes acting in the catchment, especially for arid and semi-arid regions containing temporary rivers1 [8]; iii) estimate loads when flow and concentration measurements are not continuous and simultaneous; and iv) couple the assessment of the total WF of catchment-scale crops and in-stream monitoring activities. The methodology for calibrating and quantifying nitrogen export coefficients and water balance components useful for WF assessment constitutes the first experimental work in a basin with a temporary river network. The methodology proposed here can support local authorities in the decision-making process for effective agricultural policy setting and water planning, thus fostering implementation of the EU Water Framework Directive [9].

Analysis of agricultural practices

Official data on anthropogenic activities and agricultural management practices adopted in a catchment come from agricultural censuses; however, agronomic data are generally aggregated at different spatial levels (region, province, municipality), and their downscaling needs to be verified through a detailed survey campaign. Indeed, a large difference in the amount of fertiliser used can be found between rural mountainous areas and plains areas devoted to intensive agriculture [10]. Taking into account this difference, local citizens and farmers were interviewed, and the information provided by five owners was selected as being representative of agricultural holdings for land use, cultivation techniques and management practices for the integrated census data. These owners provided information on the type, timing and amount of fertilisers used for each crop, annual crop yields, crop rotation, tillage operations and irrigation supply.

The amount of the total nitrogen (TN) application rate had to be estimated for each crop within the catchment boundaries, including the TN in synthetic fertilisers (NSF), and in animal manure (NAF) if that was used as fertiliser. In this study, the TN from NAF was estimated for each animal type, multiplying the animal-specific TN excretion rates by the live weight of each animal type [11,12] (Table 1). A distinction between indoor and outdoor farming was made. For manure produced by indoor farming, a 27.5% of TN loss during manure handling and storage was considered [13]. Appendix A contains all the acronyms/abbreviations used in this paper.

Table 1

Manure production, live weight (LW) and TN excretion rates of different animal types.

Animal type	Manure production rates	Live weight	TN excretion rate
	t t LW−1 y−1	Kg LW animal−1	kg animal−1 y−1
Horse	15.0	360.0	24.9
Dairy cattle	26.0	450.0	59.5
Beef cattle	26.0	400.0	33.6
Sheep, goat	15.0	50.0	5.0
Swine	22.0	119.5	13.9
Hen	9.5	1.4	0.3
Poultry	8.0	3.8	0.7
Rabbit	8.0	3.5	0.2

Surface-water monitoring

Hydrological processes are characterised by high temporal and spatial variability, especially in arid and semi-arid regions with temporary rivers [14,15]. A quantification of pollutants delivered to rivers requires monitoring activities that analyse all streamflow conditions (flood events, normal and low flow) [[16], [17], [18]].

In the case study location (Celone, Apulia, Italy), an automatic sampler (ISCO model 6712FS) with an internal data logger was installed [17] (Fig. 1). The sampler was connected to a flow module (ISCO 750 Area Velocity Flow Module) to measure stream water stage and velocity, which were converted to streamflow using a predefined stage-discharge rating curve. The sampler offers different sets of programming. In particular, two sets were selected: i) a time-space sampling program; and ii) a program that was triggered by water-level changes during the rising limb of the hydrograph, and flow rates during flood recession. With the first standard program, periodic samples were taken at monthly or fortnightly intervals during summer and autumn, and once or twice a week from November to June. For flood events (second program), the time intervals varied from 15 min to 2 h over the rising limb of the hydrograph, and from 2 h to one day over the flood recession. The concentrations of ammonia (N-NH4), nitrate (N-NO3), nitrite (N-NO2), total organic nitrogen (TON) and TN (TN = TON + N-NH4 + N-NO3 + N-NO2) were determined using the APAT-IRSA chemical standard analytical methods [19].

Fig. 1

Automatic sampler installed at the closing section of the Celone catchment.

In order to evaluate the contribution of point sources (i.e., wastewater treatment plants discharging into the river system), if present, it was necessary to analyse nitrogen concentrations upstream and downstream from the point-source discharge. In the Celone catchment, wastewater was treated in three treatment plants (about 3000 inhabitant equivalents) and discharged into the river (Fig. 2).

Fig. 2

Land cover map of the Celone catchment (Corine Land Cover – IV Level, 2011).

Detailed descriptions of the instruments and methods used can be found in De Girolamo et al. [10,16,17].

Riverine nitrogen export estimates

Estimating load, when flow and concentration measurements are not continuous and simultaneous, is not an easy task. In recent years, several methodologies have been developed [[20], [21], [22]]. Tan et al. [23] summarised the data requirements and applicability of the different methods. They suggested using averaging methods if no significant relationship between streamflow and nutrient concentration exists. The 'Loads' tool [1] provides monthly and annual load calculations as a result of different methods. The tool requires daily streamflow data (continuous time-series) and discrete daily concentration values.

A preliminary calculation of the daily equivalent concentration was needed for those days during which several samplings were performed (flood events). Eq. (1) was used for the Celone catchment, since the time interval of streamflow measurements was 15 min., where DailyL is the daily load (kg) passing through the river section (L), qi is measured streamflow (L) at time interval t (1, 2, …96), c is the measured or linearly-interpolated concentration (mg L−1) at time t (1, 2, …96), 0.9 is the time interval (15 × 60 = 900 s, which includes the conversion factor 1000−1). By dividing DailyL by the daily volume, the daily equivalent concentration was obtained.

For the Celone catchment, four methods, using some form of average in the calculation of the loads, we employed. These are: Method 1 – intersample mean concentration (Eq. (2)); Method 2 – intersample mean concentration using mean flow (Eq. (3)); Method 3 – linear interpolation of concentration (Eq. (3)); and Method 4 – concentration power curve fitting (Eq. (4)). We used these methods since the number of samples was high throughout the year (100 or more), and the samples covered all flow conditions (high, normal and low flow) [[20], [21], [22]].

In Method 1, the concentration values were averaged to estimate unsampled days:, where Cj is the j sample concentration and Qj is the jth flow.

Method 2 assumed that the average daily concentration on non-sampled days was determined by a simple linear average of the concentrations from the last sample data and the next sample date. The flow was assumed to be the average flow up to the next sampled concentration value., where Cj is the j sample concentration and is the average flow to the end of the j + 1 period.

Method 3 used Eq. (3), but assumed that the concentration on non-sampled days was determined by linearly interpolating between fortnightly or monthly sampled concentrations.

Meanwhile, Method 4 used a power function:, where c is the i concentration, when it exists, and aq otherwise, a being a calculated coefficient, q is i sampled discharge (flow), b is the calculated power, is the average of n concentration measurements, is the average of n discharge measurements and k is the number of time intervals in a period.

These methods provided four different values for the monthly and annual loads. The results were similar for the averaging methods (Methods 1, 2, 3; in the case study, being from 280 t y−1 and 292 t y−1), whilst Method 4 provided a lower value for load (199 t y−1). Previous studies have demonstrated that power curve methods (Method 4) underestimate high concentrations and overestimate low values [24]. The accuracy of the averaging methods depends on the number of samples and on the time between two successive samples. In temporary streams, data stratification could be applied, based on the hydrological regime, to improve the estimation.

The mean, minimum and maximum annual nitrogen riverine export (NRE,mean, NRE,min, NRE,max) were calculated in order to calibrate the leaching-runoff fractions, and estimate the uncertainty.

Nitrogen input from point sources

When data concerning point sources were missing, nitrogen input waste loads (TN point sources – NPS) were estimated by using the following equation:, where Q is the upstream flow, Q is the wastewater flow rate, Q is the flow rate downstream of the source, C is the upstream TN concentration, C is the concentration of the wastewater and C is the concentration downstream. If one sample per month is collected, the flows and concentrations are assumed to be the same throughout the month as on the day of sampling. The equation must be applied to each point source discharging into the river net.

Definition of land-use systems

Land use is the first level of information that is generally derived from a land cover map. In this study, we used the land cover map provided by the local authorities (Corine Land Cover – IV Level, 2011), which was obtained from satellite imagery (www.sit.puglia.it). Since the map has a spatial resolution of 1:100,000, and does not allow distinction among arable land types, it was reclassified by crossing data from agricultural censuses with data from local surveys and field inspections. In particular, the agricultural census provided data on crop areas at a municipal scale. Meanwhile, from local surveys and field inspections, information on crop irrigation was obtained.

Fig. 2, Fig. 3 show the initial land cover map along with the reclassified land use map obtained for the Celone catchment (Puglia Region, SE Italy), respectively. Irrigated and non-irrigated arable lands of the land cover map were divided among crucifers, durum wheat, field bean, herbage, legumes, potato, sugar beet, sunflower, tomato, vegetable crop, vetch and winter wheat, as deduced from municipal agricultural censuses and farmer interviews.

Fig. 3

Celone catchment land cover map reclassified after farmer interviews and field surveys.

In order to perform a soil-water balance, calibrate the leaching-runoff fraction and finally assess the total WF of catchment-scale crops, the watershed was preliminarily divided into land-use systems (LUSs), defined as areas with similar land use, soil characteristics and precipitation amounts [25,26]. Therefore, land-use maps, soil maps and rainfall zones (Thiessen polygons) can be intersected by means of specific GIS tools. Previously, soil maps had to be reclassified according to hydrological soil groups (i.e., A, B, C, D) [27].

Considering the Celone catchment, in addition to the precipitation gauges present in the catchment, all the gauging stations within a 25 km buffer were included in the analysis. Thus, 21 precipitation gauges were considered. Four rainfall zones, two hydrological soil groups (C, D) and 24 land-use classes have been distinguished [27] (Fig. 4). Hence, 103 LUSs were identified (Fig. 5).

Fig. 4

Land use map, soil map and rainfall zones used for identifying LUSs within the Celone catchment.

Fig. 5

Land Use Systems (LUSs) within the Celone catchment, 103 identified.

Accounting of the water footprint

The WF includes different types of water consumption, such as water volume from rainfall, which evapotranspirates (green water), irrigation water volume (blue water) and the water required to assimilate pollution (grey water).

The total WF (m3 t−1) was calculated as the sum of the green (WF), blue (WF) and grey (WF) components [28]:

To perform a complete WF assessment of crop production on a river basin scale, it is necessary to evaluate the WFs associated with each crop growing within the watershed. Hydrological and water quality models can be used for estimating the WF components, although these approaches require specific knowledge about the model and data for their calibration and validation [29].

In this study, an easy approach for estimating the green, blue and grey components of the WF assessment was performed by slightly modifying the calculation framework proposed by Franke et al. [30] and Hoekstra et al. [28] (Fig. 6). The proposed approach does not require specific modelling training.

Fig. 6

WF accounting methodology.

A calibration procedure aimed at quantifying the nitrogen export coefficients (leaching and runoff fractions) and the water balance components (runoff), by means of observed surface runoff data, measured nitrogen load and nitrogen balance results, was developed.

Green water footprint

The WF was calculated as the ratio of the volume of green water used for crop production (CWU, m3 ha−1 y−1) to the average annual crop yield, Y (t ha−1 y−1) [28]:CWU refers to the part of the precipitation that does not runoff or leach, since it is temporarily stored on top of the soil or vegetation and/or in the soil. This water can evaporate or transpire through plants, and be an important factor in agricultural production, especially in rain-fed croplands [31].

A multitude of different empirical formulae or crop models exist to estimate CWU in agriculture [28]. In this study, the ‘irrigation schedule option’ was used, since it is more accurate, and is applicable to both optimal and water-stressed growing conditions. According to this procedure, the CWU of a crop is assumed to be equal to the crop evapotranspiration under non-standard conditions (also called ‘actual’, or ‘adjusted crop’, evapotranspiration), and assuming that the soil does not receive any irrigation (ETc,adj):, where K is the single crop coefficient (dimensionless), ET is the reference evapotranspiration (mm time−1) and K is the stress coefficient (dimensionless).

Calculating evapotranspiration

The computation of ET is generally done following the methods and assumptions provided by Allen et al. [32]. A time interval of 10 days (10-d) is considered to provide a good representation of hydrological processes [33].

Since K varies in time as a function of plant growth stage, 10-d-averaged single crop coefficients (K) are calculated for each crop in the catchment from the crop coefficient curves, which are constructed using initial K (K), middle K (K) and end Kc (K). Data concerning crop planting dates and length of cropping season were provided from the above-mentioned interviews with farmers and local dealers.

Referring to the calculation methodology adopted for the ET estimate (mm d−1), the following equation [34] was applied to the temperature gauges, if solar radiation, relative humidity and wind speed data, required by other methods (i.e., the Penman–Monteith equation), were missing in the study area [32,35]:, where is the latent heat of vapourisation (MJ kg−1), RA is the extraterrestrial solar radiation (MJ m−2 d−1) and Tmax, Tmin and T are the daily maximum, minimum and mean air temperatures (°C), respectively.

Daily is obtained by applying the Harrison formula [36]:

Daily ET values were calculated by applying Eq. (9), then appropriately summing these in order to obtain values on a 10-d basis (ET).

Finally, a GIS-based inverse distance weighted method was used in order to spatially interpolate the punctual ET values in the entire watershed, considering a 10-d time-interval [37,38]. For the Celone catchment, in addition to the temperature gauges present in the catchment, all the gauging stations within a 25 km buffer were included in the analysis. Thus, 13 temperature gauges were considered.

After calculating the crop evapotranspiration under standard conditions (ET), K was evaluated as follows [32]:, where TAW (mm) is the total available water in the root zone, RAW (mm) is the readily available water in the root zone and D (mm) is the root zone depletion at the start of the 10-d period considered. Formulae (12) and (13) were used to assess TAW and RAW values, respectively:, where is the water content at field capacity (m3 m−3), is the water content at wilting point (m3 m−3), Z is the rooting depth (m) and p is the soil-water depletion fraction for no stress, the values of which have been tabulated by Allen et al. [32].

The and depend on the type of soil, and average values can be estimated with the software Soil Water Characteristics, implemented by the USDA Agricultural Research Service. Z is estimated considering the lowest value between the depth of the soil layers in the watershed and that reported for various crops by Allen et al. [32].

Soil-water balance computation

Lastly, a water balance computation for the root zone was implemented on a 10-d basis, in order to estimate root zone depletion at the end of the 10-d period (D, mm). Hence, a GIS model with a resolution of 20 m was developed. According to Allen et al. [32], the incoming (irrigation, rainfall) and outgoing (runoff, deep percolation, evapotranspiration) water flux into the crop root zone have to be assessed. Water transferred horizontally by subsurface flow in or out of the root zone is ignored. Moreover, if the groundwater table is more than ∼1 m below the bottom of the root zone, the amount of water transported upwards by capillary action can be assumed to be zero.

Therefore, the following equation for the water balance was used:, where P (mm) is the net precipitation, RO (mm) is the runoff from the soil surface, I (mm) is the irrigation depth, ET (mm) is the actual crop evapotranspiration and DP (mm) is the water loss out of the root zone by deep percolation. D and D can assume values between 0 and TAW.

The P is determined as follows:, where P (mm) is the total precipitation amount. Thiessen polygons are built and rainfall zones are distinguished [25,26]. ET is determined according to Eq. (8), whilst Eq. (16) is used for the determination of DP:The I values are set to zero for irrigated crops, both in Eqs. (14) and (16), in order to estimate CWU [28,39].

RO was estimated using the Soil Conservation Service Curve Number (SCS-CN) method [27], which is one of the most commonly-used models due to its simplicity and the requirement for few data [40,41]. This model is used to predict the depth of surface runoff (RO, mm) for a given rainfall event, and can be expressed as follows:, where S is the potential maximum retention or infiltration (mm) and P is the total storm rainfall (mm). S was evaluated using the following equation:, where CN is the curve number (dimensionless) that ranges from 1 (minimum runoff) to 100 (maximum runoff). This parameter was determined and tabulated based on hydrological soil group and soil cover type, treatment and hydrological condition [42]. The tabulated values (CN II) refer to the average antecedent moisture condition (AMC II). The AMC definition depends on the total 5-d antecedent rainfall, and the season category (dormant or growing) that is defined from daily average temperatures [43]. Different CN conversion formulae, from AMC II, to dry AMC (AMC I – CN I) and wet AMC (AMC III – CN III), have been proposed [44]. For this case study, the Hawkins et al. [45] CN conversion formulae were used.

In summary, CN II tabulated values were associated with each LUS identified in the basin; AMC was evaluated for the identified rainfall zones and, where necessary, CN I and CN III were calculated. After that, Eqs. (17) and (18) were applied, and the runoff associated with single precipitation events was estimated for each LUS. Considering a 10-d time interval, the runoff was appropriately added, and the preliminary RO (mm) was obtained. This latter value was then modified, following the calibration procedure described below, and based on the continuous flow measurements at the gauge.

The water balance for the root zone (Eq. (14)) was initiated in the first 10-d period of a really wet month. Therefore, it was assumed that, in that month, the root zone was near field capacity and, hence, D = 0, K = 1 and ET = ET [46].

Calibrating curve numbers

Daily mean baseflow (BF, m3 s−1), daily mean interflow (IF, m3 s−1) and total daily mean wastewater discharge (WW, m3 s−1) were subtracted from the mean daily streamflow recorded at the gauge (Q, m3 s−1), as follows:, where SF (m3 s−1) is the estimated daily mean stormflow, Q is obtained from the continuous measurement of flow, and BF and IF can be assessed by means of the baseflow filter program (swat.tamu.edu/software/baseflow-filter-program). Fig. 7 shows the measured streamflow, the sum of the daily mean baseflow and the daily mean interflow in the Celone catchment closing section.

Fig. 7

Measured daily streamflow (Q) and sum of baseflow (BF) and interflow (IF) in the Celone catchment closing section.

Volumes of surface runoff (SF, m3) were estimated throughout the study period, considering a 10-d time step (i). Based on these values, the CNs associated with each LUS were recalculated, so that the sum (m3) of RO (, Eq. (17)) was equal to SF (Eq. (19), the target function). To do this, a spreadsheet was specifically created, and the target function was set. Finally, calibrated CN values were used to estimate the RO (mm), required for the soil-water balance.

Results for the Celone catchment can be found in D’Ambrosio et al. [47].

Blue water footprint

The WF refers to the consumption of groundwater and/or surface-water resources that are utilised in crop production (i.e., irrigation water) [28]. In other words, it is the amount of groundwater and/or surface water that does not return to the source in the form of return flow, and is different from water withdrawn for irrigation, insofar as this water is returned to where it came from.

The WF (m3 t−1) was calculated by dividing the total volume of blue water used – CWU (m3 ha−1 y−1) – by the quantity of the annual production – Y (t ha−1 y−1):The CWU (mm time−1) was calculated by performing another soil-water balance (Eq. (9)) on a 10-d basis, and irrigation (I) was considered, as proposed by Hoekstra et al. [28] and applied by Mekonnen and Hoekstra [39], de Miguel et al. [48], Zhuo et al. [46] and Rulli and D’Odorico [49]. I values could be deduced from the above-mentioned interviews. The following equation was then used:, where is the adjusted crop evapotranspiration, estimated by means of the same procedure applied for ET evaluation (Eq. (8)), but considering also I in Eqs. (14) and (16). In the case of rain-fed crops, CWU is zero.

Grey water footprint

The WF refers to the volume of water needed to dilute a load of pollutants discharged into the natural water body in such a way that the quality of the receiving water system is not compromised, with respect to specific quality standards and natural background concentrations [6,28]. In this study, the WF related to nitrogen use was quantified, thus excluding the effects of other nutrients and fertilisers. Hence, the intensity of water pollution caused by agricultural activities and, in particular, by the TN application rate, was measured.

The WF (m3 t−1) was calculated by dividing the dilution water requirement, CWU (m3 ha−1 y−1), by the crop yield, Y (t ha−1 y−1) [28]:The CWU (mm y−1) was calculated by multiplying the fraction of TN that leaches or runs off (leaching-runoff fraction – α) by the TN application rate (AR, kg ha−1 y−1), and dividing this by the difference between the maximum (C) and natural (C) concentration (mg l−1) of TN in freshwater:

The AR was estimated for each LUS, based on the above-mentioned interviews. Meanwhile, a calibration procedure was defined in this study in order to divide α between leaching (α) and runoff (α), as follows. Finally, CWU was estimated by summing the values associated with runoff (CWU) and leaching (CWU).

Calibrating the runoff and leaching nitrogen fraction

In most of the previous studies, α has been set at a constant value of 10% [39,46,50] or 7% [51]. In contrast to the use of a static α throughout the watershed, the procedure suggested by Franke et al. [30], and applied by Brueck and Lammel [52], Munro et al. [53] and Gil et al. [54], was preliminarily used in this study. This approach considers that α depends on potential factors – (j) – which are atmospheric input (TN deposition), soil type (texture and natural drainage), climate (precipitation) and agricultural practice (TN fixation, application rate, plant uptake and management practice). The α values were calculated for each LUS (k), using the following equation:, where α is the minimum leaching runoff fraction (0.01), α is the maximum leaching runoff fraction (0.25), s is the score for the above-mentioned potential factor, j, associated with the LUS k, and w is the weight of the factor j associated with k.

Franke et al. [30] provided specific criteria used to score (s) and weight (w) all the different influencing factors (j). Unlike in previous studies, the procedure suggested by Franke et al. [30] was slightly modified in this study, and α was divided between leaching (α) and runoff (α). Then, a zero weight (w) was assigned to factors specifically related to runoff (i.e., texture and natural drainage relevant to runoff) and leaching (i.e., texture and natural drainage relevant to leaching), respectively.

At the watershed closing section, TN runoff (R, kg y−1) and TN in soil and leaching (L, kg y−1) were estimated using Eqs. (25) and (26):, where n is the number of LUSs identified within the watershed, α and α are the runoff and leaching fraction associated with the LUS k, AR (kg ha−1 y−1) is the application rate associated with the LUS k and A (ha) is the surface of the LUS k.

Following this, α and α were recalculated, following a calibration procedure based on the field measurements [55]. Thus, R (Eq. (25)) and L (Eq. (26)) were equalised to TN runoff (R, Eq. (27)) and TN in soil and leaching (L, Eq. (28)) estimated from field measurements. Hence, the values of α and α were multiplied by an appropriate constant factor. A spreadsheet was specifically created in order to set the target function and obtain the constant factor., where N (kg y−1) is the annual TN riverine export, N (kg y−1) is the TN biological fixation, N (kg y−1) is the TN atmospheric deposition, N (kg y−1) is the TN in wastewater, N (kg y−1) is the TN naturally present in the river, Δ (kg y−1) is the difference between TN input (N) and output (crop uptake – N, NH3 volatilisation – N and denitrification in soil – N) in the study area. Moreover, N and N are N and N, respectively, associated with soil of hydrological group D, where infiltration does not occur.

N was assessed by multiplying C for the total discharge, Q (m3 y−1), measured at the gauge, and N was determined, as described in the section above. Meanwhile, Δ was obtained from a nitrogen balance [10,56,57]. In particular, for the Celone catchment, N (64,891 kg y−1), N (39,744 kg y−1) and Δ (306,397 kg y−1) values are reported in De Girolamo et al. [10].

The values of L were set to be greater than zero. If the uncertainty in the WF estimate related to α variability was assessed, the mean, minimum and maximum annual nitrogen riverine export (N) were considered as N in Eqs. (27) and (28).

Results for the Celone catchment can be found in D’Ambrosio et al. [47].

Maximum and natural TN concentrations in surface and groundwater

Regarding the C value in Eq. (23), despite that the idea of measuring water pollution in terms of the amount of water needed to dilute pollutants can be traced back to Falkenmark and Lindh [58], and was continued by Postel et al. [59] and Chapagain et al. [50], still today there are uncertainties related to the standardisation of water-quality standards that should be used for a consistent WF assessment, taking into account the diverse ambient water quality and aquatic ecosystems, as well as the presence of several pollutants in water-bodies [6]. Generally, WF assessments have used drinking-water standards. Regardless of the fact that this value is referred to a surface-water or groundwater body, the US EPA (10 mg N-NO3 l−1) or the European Union/World Health Organization (50 mg NO3 l−1, i.e., 11.3 mg N-NO3 l−1) nitrogen standards for drinking-water are the most commonly used water-quality standards [39,50,51,[60], [61], [62]]. Also, 50 mg NO3 l−1 is the maximum concentration permitted by the EU Nitrate Directive in groundwater [63]. In the literature, only a few studies have used ambient water-quality standards [46,64,65]. In Italy, the Decree of the Ministry of the Environment n. 260/2010 [66] identifies, among various physicochemical factors, the threshold concentrations of NH4, NO3 and NO2 that are required to support a functioning ecosystem. Concerning groundwater, a good chemical status is reached if the concentrations of NH4, NO3 and NO2 are lower than 0.5 mg l−1 (i.e., 0.4 mg N-NH4 l−1), 50 mg l−1 and 0.5 mg l−1 (i.e., 0.1 mg N-NO2 l−1), respectively. Meanwhile, the threshold values associated with the good water-quality status of a surface-water body for N-NH4 and N-NO3 are 0.06 and 1.2 mg l−1, respectively. Currently, Italian legislation does not provide ambient quality thresholds for TN in either surface water or groundwater. Following Liu et al. [6], the TN ambient water-quality standard (good) adopted for Celone catchment surface water is 3 mg l−1. Meanwhile, the standard adopted for groundwater is 4.6 mg l−1 [67].

Regarding the C value in Eq. (23), many previous studies have considered this value to be equal to zero, due to a lack of data [49,50,61,63,68]; however, such an assumption leads to an underestimation of CWU because C is generally higher than zero. For the Celone catchment, since local data are not available, the C of TN was set 0.4 mg N l−1 in both river water and groundwater, as recommended by Franke et al. [30], and used by Mekonnen and Hoekstra [69] and Liu et al. [6].

Limitations of the study

The main limitations of the procedure adopted in this study for the calibration of runoff and the nitrogen leaching runoff fraction are [47]:

the SCS-CN method was used beyond its original scope, since it was not applied considering a single storm event, but the sum of storm events that happened in a 10-d period. In addition, small events that produce no runoff were not excluded. These limitations can lead to an overestimation of CN II values;

the limited time period analysed (1 year), with wetter than average conditions, could have greatly influenced the outcomes, especially the WF that could have resulted in being higher than average. An average over multiple years would be more appropriate for representing the WF and the general status of water scarcity;

the discretisation of the watershed into LUSs (based on land use, soil type and Thiessen polygons) adopted in this study leaves out local factors, such as slope, rainfall intensity and river distance. The latter factors influence hydrological processes and the leaching and runoff fraction, hence, neglecting them could lead to an overestimate or underestimate of these variables;

the high fragmentation of land use, and the different management practices adopted within the catchment, makes the assignment of actual agricultural practices to each field impossible, hence, the agricultural practices used in the calculations (i.e., fertiliser amount, crop yield) can be affected by certain uncertainty;

TN accumulation and degradation processes in the receiving water-bodies were neglected in the TN load calculations, whilst biochemical processes can be relevant, especially in temporary rivers;

the in-stream monitoring of TN concentration program (i.e., number of samples, time between two successive samples, hydrological conditions), as well as the method used for estimating load, have a great influence on load estimation that results in a large uncertainty in the runoff and leaching nitrogen fraction estimation. A standard procedure is needed for fixing Cmax and Cnat. Maximum allowable concentrations fixed on standards for drinking-water lead to an underestimation of grey water, as well as a natural background fixed as equal to zero.

Funding

Monitoring activities were funded by the MIRAGE Project (contract 211735, 7th EU Framework Programme 2007–2011).

Subject area	• Agricultural and Biological Sciences
More specific subject area	Water Footprints assessments
Method name	Calibrating the water footprint accounting of crops with in-stream measurements at catchment-scale
Name and reference of original method	• Water footprint accounting for catchments and river basins. Calculation of the green, blue and grey water footprint of growing a crop or tree - Irrigation schedule option (Hoekstra, A.Y., Chapagain, A.K., Aldaya, M.M., Mekonnen, M.M., 2011. The Water Footprint Assessment Manual. London – Washington DC) • Estimate the leaching-runoff fraction for nitrogen diffuse pollution sources (Franke, N.A., Boyacioglu, H., Hoekstra, A.Y., 2013. Grey water footprint accounting: Tier 1 supporting guidelines. Unesco-IHE, Delft)
Resource availability	- Software: Soil Water Characteristics (SPAW) (ars.usda.gov/research/software) - Software: Baseflow Filter Program (swat.tamu.edu/software/baseflow-filter-program) - Loads tool [1]

Abbreviation	Description	Unit of measure
TN	Total Nitrogen	–
Ak	Surface of the Land Use System k	ha
AR	TN application rate	kg ha−1 y−1
ARk	TN application rate associated with the Land Use System k	kg ha−1 y−1
BF	Daily mean baseflow	m3 s−1
Cmax	Maximum concentration of TN in the water bodies	mg l−1
CN	Curve number	dimensionless
Cnat	Natural concentration of TN in the water bodies	mg l−1
CWUblue	Blue crop water use	m3 ha−1 y−1 (mm time−1)
CWUgreen	Green crop water use	m3 ha−1 y−1 (mm time−1)
CWUgrey	Dilution water requirement	m3 ha−1 y−1 (mm time−1)
CWUgrey,L	Dilution water requirement (leaching)	m3 ha−1 y−1 (mm time−1)
CWUgrey,R	Dilution water requirement (runoff)	m3 ha−1 y−1 (mm time−1)
DPi	Water loss out of the root zone by deep percolation at the end of the 10-d period	mm (10-d)−1
Dr,i	Root zone depletion at the end of the 10-d period	mm (10-d)−1
Dr,i-1	Root zone depletion at the start of the 10-d period	mm (10-d)−1
ET0	Reference evapotranspiration	mm time−1
ET0,i	10-d reference evapotranspiration	mm (10-d)−1
ETc,adj	Actual (or adjusted) crop evapotranspiration assuming that the soil does not receive any irrigation	mm time−1
ETc,adj,i	Actual crop evapotranspiration at the end of the 10-d period assuming that the soil does not receive any irrigation	mm (10-d)−1
ETc,adj,iI≠0	Actual crop evapotranspiration at the end of the 10-d period considering irrigation	mm (10-d)−1
ETc,i	10-d crop evapotranspiration under standard conditions	mm (10-d)−1
IF	Daily mean interflow	m3 s−1
Ii	Irrigation depth at the end of the 10-d period	mm (10-d)−1
Kc	Single crop coefficient	dimensionless
Kc,end	End single crop coefficient	dimensionless
Kc,i	10-d average single crop coefficient	dimensionless
Kc,ini	Initial single crop coefficient	dimensionless
Kc,mid	Middle single crop coefficient	dimensionless
Ks	Stress coefficient	dimensionless
Ks,i	10-d stress coefficient	dimensionless
L	TN in soil and leaching calculated with Eq. (26)	kg y−1
LM	TN in soil and leaching estimated for the study area with Eq. (28)	kg y−1
NAD	TN atmospheric deposition	kg y−1
NAD,soil D	TN atmospheric deposition associated with soil D	kg y−1
NAF	TN in animal manure	kg y-1
NBF	TN biological fixation	kg y−1
NBF,soil D	TN biological fixation associated with soil D	kg y−1
NCU	TN crop uptake	kg y−1
NNAT	TN naturally present in the river	kg y−1
NPS	TN point sources	kg y−1
NRE	TN riverine export (NRE,mean; NRE,min; NRE,max)	kg y−1
NRE,max	Annual maximum TN riverine export	kg y−1
NRE,mean	Annual mean TN riverine export	kg y−1
NRE,min	Annual minimum nitrogen riverine export	kg y−1
NSF	TN in synthetic fertilisers	kg y−1
NV	NH3 volatilisation	kg y−1
p	Soil water depletion fraction for no stress	dimensionless
P	Total storm rainfall	mm time−1
Pi	Total precipitation amount at the end of the 10-d period	mm (10-d)−1
Pn,i	Net precipitation at the end of the 10-d period	mm (10-d)−1
QG	Mean daily streamflow recorded at the gauge	m3 s−1
QTOT	Total discharge measured at the gauge	m3 y−1
R	TN runoff calculated at the watershed closing section with Eq. (25)	kg y−1
RA	Daily extraterrestrial solar radiation	MJ m−2 d-1
RAW	Readily available water in the root zone	mm
RM	TN runoff estimated at the watershed closing section with Eq. (27)	kg y−1
ROi	Runoff from the soil surface at the end of the 10-d period	mm
S	Potential maximum retention or infiltration	mm
SFG	Estimated daily mean stormflow	m3 s−1
SFG,i	Volumes of surface runoff at the end of the 10-d period	m3 (10-d)−1
sj,k	Score for the potential factor that influence leaching and runoff (j) associated with the Land Use System k	dimensionless
TAW	total available water in the root zone	mm
Tmax	Daily maximum air temperature	°C
Tmean	Daily mean air temperatures	°C
Tmin	Daily minimum air temperature	°C
WF	Total water footprint	m3 t−1
WFblue	Blue water footprint	m3 t−1
WFgreen	Green water footprint	m3 t−1
WFgrey	Grey water footprint	m3 t−1
wj,k	Weight of the potential factor that influence TN leaching and runoff (j) associated with the Land Use System k	dimensionless
WW	Total daily mean wastewater treatment plant (WWTP) discharge	m3 s−1
Y	Average annual crop yield produced	t ha−1 y−1
Zr	Rooting depth	m
α	Leaching-runoff fraction	dimensionless
αk	α values calculated for each Land Use System k	dimensionless
αL	Leaching fraction	dimensionless
αL,k	Leaching fraction associated with the Land Use System k	dimensionless
αmax	Maximum leaching-runoff fraction	dimensionless
αmin	Minimum leaching-runoff fraction	dimensionless
αR	Runoff fraction	dimensionless
αR,k	Runoff fraction associated with the Land Use System k	dimensionless
Δ	Difference between TN input and TN output	kg y−1
θFC	Water content at field capacity	m3m−3
θWP	Water content at wilting point	m3m−3
λ	Latent heat of vaporisation	MJ kg−1