Hydro-economic modeling of water scarcity under global change: an application to the Gállego river basin (Spain)

Abstract

Integrated approaches are needed to assess the effects of global changes on the future state of water resources at regional scales. We develop a hydro-economic model of the Gállego catchment, Spain, to assess how global change and policy options affect the catchment’s water scarcity and the economic implications to the agricultural sector. The model couples physical processes (hydrology) and regulatory and economic processes (agricultural water demand, reservoir operation). Five scenarios, covering currently ongoing changes in climatic conditions, agriculture and hydrological planning, are evaluated. Our results suggest that the scenarios’ impacts on water resources and regional agricultural income are significant. Policy responses such as investments in modernization of irrigation technology would mitigate the negative impacts of climatic change on the agricultural sector, but the implementation costs outweigh the extra regional agricultural income. Also, a planned reservoir extension project appears ineffective, even considering effects of climatic change. Although our results are site-specific, our methodology is relevant to other areas that face comparable problems of water scarcity.

Appendices

Appendix 1: Details on the GEOTRANSF model (hydrology)

Daily precipitation and temperature at 32 rain gauges and 26 temperature stations within the catchment were available from CHE (2007) during the period 2001–2005. Daily precipitation was distributed spatially by using the Ordinary Kriging method and considering the eight nearest stations to the interpolation point, which in our scheme corresponds to the center of mass of each subcatchment. Temperature was computed for each subcatchment by linear interpolation with elevation. The interpolation was performed by using the six closest temperature stations, and the subcatchment was assigned the temperature corresponding to its mean elevation. Potential evapotranspiration has been calculated with the method of Hargreaves and Samani (1982) which depends only on the mean, maximum and minimum daily temperatures of the subcatchments. Notice that potential evapotranspiration was transformed into actual evapotranspiration by means of a linearly varying water stress reduction function following the approach proposed by Rodríguez-Iturbe and Porporato (2004, chap. 2.1.5). Land use and geology information available for the Gállego catchment have been used in order to apply the Soil Conservation Service Curve Number model (SCS-CN, Michel et al. 2005) which is the method present in the infiltration module of GEOTRANSF.

Calibration and validation

The classical Nash–Sutcliffe (NS) efficiency coefficient (Nash and Sutcliffe 1970) was adopted to evaluate the performances of hydrological modeling. According to this metric, NS equals 1 when the match between the model and the observed data is perfect, while when NS < 0 the model should be disregarded because its predictions are worse than approximating all the observational data with their sample mean. As an example, with reference to the Saragoza node, the set of parameters obtained by calibrating the model to the year 2003 produced a relatively high NS value of 0.81 and NS ranging from 0.51 to 0.8 when applied in the validation of the remaining years. This is evidence that the set of parameters obtained by calibrating to the year 2003 provides a good reproduction of the streamflow recorded in other periods as also shown in Fig. 8. In particular, the model has performed well and has been able to predict correctly the magnitude and timing of storm events (see Fig. 8) and also the shape of the stormflow recessions both in dry and in wet periods, although peak flows are sometimes underestimated. Notice that similar results were also obtained for the other nodes.

Fig. 8

Observed (blue line) versus simulated (pink line) daily streamflow during the period 2001–2005 computed at node 6 and obtained by calibrating the model on the year 2003 (color figure online)

A more detailed description of the implementation of GEOTRANSF model to the Gállego catchment and of the calibration and validation procedure can be found in the work of Majone et al. (2012). It presents an assessment of the projected impacts of climatic change on streamflow and water resources of the catchment by adopting an ensemble of future climate experiments (one of which is used in the present work) and provides a robust validation of the whole modeling framework (i.e., hydrological modeling and reservoir operation model) by also verifying the model performances during a much longer time period, that is, 1961–1990.

Appendix 2: Details to the economic model

Model

The agro-economic model maximizes regional profits Π as follows (for ease of notation, a subscript to denote years is not included):

$$\Uppi = \sum\limits_{n \in N} {\sum\limits_{g \in G} {\left( {q_{ng} \cdot p_{g} } \right) - \left( {w_{ng} \cdot r} \right)-t_{ng} } + s_{ng} } $$

(1)

with $q_{ng} =\sum\nolimits_{j \in G} {x_{nj} \cdot y_{j} } \; , \; \; w_{ng}= \sum\nolimits_{m \in M} {\sum\nolimits_{j \in G} {x_{nj} \cdot u_{jm} } } ,$ and $t_{ng} =\sum\nolimits_{j \in G} {x_{nj} \cdot c_{g} }$, where q _ng  ≥ 0 is the production of crop g ∈ G in municipality n ∈ N, p _g  ≥ 0 is the crop price, w _ng  ≥ 0 is water use, r is the water price, t _ng is the total remaining production costs, and s _ng is a subsidy. x _nj is the area cultivated with Land use system j ∈ J (defined by crop g and irrigation technique i) in municipality n, y _j  ≥ 0 is the yield, u _jm  ≥ 0 is the water use of Land use system j in month m ∈ M, and c _g  ≥ 0 are crop-specific costs. The following constraints hold:

$${\sum\limits_{j \in I} {x_{nj} } \le l_{ni} } $$

(2)

$$\sum\limits_{n \in N} {\sum\limits_{g \in G} {w_{ng} } \le a} $$

(3)

$$\gamma_{ng} = {\sum\limits_{z \in Z} {\gamma_{ngz} b_{z} } } $$

(4)

Constraint (2) is a land constraint, saying that the total cultivated area in municipality n should not outweigh total available land l _n in municipality n, subdivided by irrigation type i ∈ I. Water constraint (3) says that water use over all municipalities should not outweigh total water availability a. Constraint (4) assures a convex combination of historical crop mixes, where γ _ng denotes current crop choice in municipality n, γ _ngz denotes the historical crop mix in year z, and b _z  ≥ 0 is a variable that denotes the share of each historical crop mix in the solution with $\sum\nolimits_{z \in Z}^{{}} {b_{z} = 1}$. This constraint assures that the solution is a convex combination of historical crop mixes, in the sense that it prevents the model from choosing crop mixes that lay beyond the boundaries of crop mixes that have been chosen in the past. Doing so, this constraint prevents unrealistic crop mixes, thereby implicitly accounting for crop rotation and other practices that affect crop choice (McCarl 1982; Chen and Önal 2012). Because the reference years have a large variation in terms of water availability, this constraint allows the model to provide realistic output in simulated drought years caused by climatic change. The drawback of this approach is that an aggregation bias might occur, in the sense that the model is not capable of producing sensible crop choices when water availability (or some other model input) has a value outside the range of its historical levels. In our results, we did not find strong effects of this drawback, mainly because of the large variation in water availability in the reference period, and that changes implied by climatic change are largely within this range.

An extension to our economic model allows estimating salt emissions from agricultural activities. Salt levels in irrigation return flows are closely related to cropping patterns, irrigation techniques and management, irrigation time and period, soil characteristics, land levelling, annual climatic characteristics and adaptation to climatic conditions. These are roughly the same factors that contribute to irrigation efficiency at the field level (e.g., Causapé et al. 2006). We do so by using salt emission functions and parameter values based on Esteban and Albiac (2007). These salt emission functions calculate the salt emission e _n in municipality n as a function of water use:

$$e_{n} = \sum\limits_{g \in G} {\beta_{0} + \beta_{1} w_{ng} }$$

Esteban and Albiac (2007) calibrated β ₀ and β ₁ on the neighboring Bardenas area, and we corrected these values for differences in irrigation efficiency and techniques.⁸ We remark that a specific transport module for salt emissions is not included in the hydrological model and thus the effect of downstream reuse of water with high concentrations is not considered.

Data

Crop price data, cost data, subsidy data and yield data are taken from Mema (2006) for the year 2005 (see Table 3). Water price data are taken from Esteban and Albiac (2007). Water use data, irrigation technique data, land availability data and historical crop mix data are obtained from the research institute Centro de Investigación y Tecnología Agroalimentaria de Aragón (CITA), Spain. Finally, water availability is calculated on irrigation district level using a database that registers the water outflow from the La Sotonera dam (CHE Web site).

Table 3 Data used in the economic model: prices and costs of the six crops included in the model