Assessing the impact of climate change on water resources, crop production and land degradation in a semi-arid river basin

Semi-arid regions suffer from shortage of water. Climate variation alongside anthropogenic factors such as population growth, land use changes, and overuse of water resources especially for irrigation, are significant concerns. Climate change and its impact on water resources are a further challenge in the sensitive and fragile ecosystems such as the semi-arid regions of Iran. The availability of water resources, crop production and the process of land degradation should be future considerations in this area in order to build up a sustainable ecosystem and to improve food security. However, the uncertainty of global circulation models (GCMs) may lead to different results in the same area (Singh & Bengtsson 2005; WWF 2010; Gosain et al. 2011). Therefore, we developed an ensemble of GCMs to predict the water availability and also the capacity of crop production in the semi-arid region of Iran.

Hijioka et al. (2014) in IPCC-AR5 mentioned that increasing water demand and mismanagement of water resources give rise to water scarcity as a major challenge for most regions in Asia. According to IPCC (2007a), climate change has a pernicious effect on water resources and freshwater availability in most of the regions of the world. It can also have a significant impact on the hydrological cycle (Piao et al. 2009; Wu et al. 2012). IPCC (2013) in their AR5 report revealed that human influence in global warming and changing of water cycle has grown since the fourth assessment report (AR4). The quantum of damages, however, can differ from one place to another. The variation of precipitation and temperature characteristics can lead to land degradation especially in fragile arid and semi-arid ecosystems (Meadows & Hoffman 2003). The variation of intensity and variability of precipitation leads to increasing risks of droughts and floods in most regions. Furthermore, alteration in floods and droughts and a rise in water temperature are changing the water quality (IPCC 2007b). IPCC (2014) reported that drought frequency would likely rise in the current dry regions by the end of the 21st century. Shifts in precipitation and temperature result in variations of groundwater recharge and also cause water table fluctuations (Changnon et al. 1988; Zektser & Loaiciga 1993). In addition, changes in groundwater recharge can have an influence on evapotranspiration (ET), groundwater flow direction, groundwater level and surface water–groundwater interaction (Ali et al. 2012).

Land use changes (particularly urbanization) could alter the hydrological response of an area by changing the amount of rainwater that goes into surface discharge and groundwater recharge (Baker & Miller 2013). However, the impact of land use/land cover change varies with the climatic conditions (Kim et al. 2013). Human activities and climate variability can influence the hydrological processes (e.g., annual runoff, stream flow) especially in semi-arid regions (Zhan et al. 2013; Xu et al. 2013). Zhou et al. (2013) revealed that surface runoff and base flow are very sensitive to urbanization, whereas ET and annual streamflow are less sensitive. Many previous studies identified a combined effect of climate and land use change on water components (e.g., surface runoff) (Tu 2009; Tong et al. 2012; Kim et al. 2013).

Food-producing capacity and livestock production can be significantly affected by climate change for years to come. While some areas may experience a decrease in crop production, others are likely to increase. CO₂ concentration and temperature are two important factors affecting crop production. While increasing CO₂ concentration as the main driver of climate change could raise production of some crops (e.g., wheat), the changing climate, in general, is likely to have a negative effect on the length and quality of the growing season. In addition, having a higher intensity of droughts and floods could have countless consequences for crop production. On the other hand, a temperature increase of a few degrees is expected to generally raise crop production in temperate areas; however, greater warming may decrease crop yields (Adams et al. 1998; Raleigh & Urdal 2007).

Assessing the impact of climate change on hydrological processes can be used with various hydrological models (e.g., Rosenberg et al. 1999; Kirshen 2002). For instance, the hydrologic unit model of the USA was used to assess the impact of global warming on the hydrology of the Ogallala area in the United States, where the comparison of different GCMs shows an influence on the reduction of groundwater recharge (Rosenberg et al. 1999). Eckhardt & Ulbrich (2003) used a revised version of soil and water assessment tools (SWAT) to investigate the impact of climate change on groundwater recharge and stream flow in a central European low mountain range. The results of their studies show little effect of climate change on mean annual groundwater recharge and stream flow. In this research, we used the SWAT eco-hydrologic model (Arnold et al. 1998) to study the impact of climate change on water resources and crop productions. Various investigations show the benefit of SWAT to assess the impact of climate change (Fontaine et al. 2001; Young et al. 2009; Faramarzi et al. 2013).

As water resources in Iran become scarcer due to recurrent droughts and rising demand, accurate knowledge of the available resources in the future is mandatory for successful management. The fresh water availability in the semi-arid region of Razan–Ghahavand has experienced severe reduction due to increasing demand for sanitation, drinking water, manufacturing and agriculture. All of these factors trigger the expansion of land degradation in this area. To improve water resource management, water use efficiency, sustainability of agriculture and land management to combat degradation in the future, precise knowledge of water resources is needed. There is a close relationship between water resources, industry, agriculture and urban dynamics in this area. Therefore, it is essential to study the climate change impact on hydrology and water resources.

The main objective of this research is to investigate the impact of climate change on water resources' components such as groundwater recharge, soil water content, surface runoff and actual evapotranspiration (AET) for 2046–2065. The effect of land use change (e.g., urbanization) on surface runoff, estimation of crop yields and an analysis of future land degradation are the other objectives of this research. We used the ensemble of four global climate models to quantify the impact of climate change on water resources and crop production in the mid-21st century.

The study area called Razan–Ghahavand is located in the central drainage basin of Iran. The area is approximately 3,100 km². The range of the highest and lowest elevation inside the watershed is roughly 1,265 m. A variety of species has evolved from the highest to the lowest elevation following this gradient. The climate of this region is semi-arid with an annual average temperature around 11 °C and annual average precipitation of about 295 mm. The main river in the watershed, called Gharehchay, enters the watershed from the west whereas the watershed's outlet is controlled by the Omarabad hydrometric station in the eastern part of the watershed. In addition, two other streams, Sirab Khomigan and Zehtaran, respectively, join the main river from the northern part of the basin. In Table 1, the watershed area covered by the single hydrometric stations and their sub-basins is shown. According to the land use map, about 52% of the area is grazing land covered by species that occur in different numbers and with a varying capacity for livestock rearing. Irrigation farming covers 26% and rain-fed cultivation covers 5% of the watershed area.

We used SWAT (Arnold et al. 1998) to assess the influence of climate change on water resources and crop production. SWAT is a hydrological model to study the quality and quantity of surface and ground water resources and predict the impact of land use, land management and climate change on water resources. It is a physical, distributed and continuous time model that operates on a daily time assessment. The main components of SWAT include hydrology, climate, nutrient cycle, sediment movement, crop growth and agricultural management. Hydrological processes in SWAT simulate surface runoff, potential evapotranspiration (PET), percolation, lateral subsurface flow, groundwater flow to streams from shallow aquifers, snowmelt, transmission losses from streams, and water storage and losses from ponds (Neitsch et al. 2011). Hydrological response units (HRUs) including homogenous land use, soil and slope characteristics are the units of water balance calculations. The water in each HRU is stored in four storage volumes including snow, soil profile (0–2 m), shallow aquifer (2–20 m) and deep aquifer (>20 m).

Basic input data include climate data, soil information, a digital elevation map, a river map, a land use classification plus crop and agricultural management data. All data are needed to set up a SWAT hydrologic model run. Climate data consist of daily precipitation, daily minimum and maximum temperature and daily solar radiation obtained from the weather service of the Iranian meteorological organization from 1977 to 2008. As the watershed is not located at the headwater of a particular basin, the inlet was defined for catchment. In other words, inlet is defined as estimating the amount of water entering the basin. The surface runoff inside the catchment is estimated by the SCS curve number method based on daily precipitation, soil hydrologic groups, antecedent soil moisture plus different mapped land use. To adapt the model to the specific study area's conditions, the curve number (CN) is modified in relation to changing slope factors. To define the impact of elevation difference on the model, elevation bands are defined for sub-basins, an elevation difference of more than 100 m using a specific temperature lapse rate [°C/km] and precipitation lapse rate [mm H₂O/km] value. To route water through the channel network, a variable storage routing method is chosen. The crop growth is simulated based on the Erosion-Productivity Impact Calculator (EPIC) method (Williams et al. 1984) by determining leaf area development (LAI), light interception and radiation use efficiency. Winter wheat is selected as the dominant crop in agricultural lands both for rain-fed and irrigated lands. Schedule planning (e.g., time of planting, irrigation, fertilization, harvesting) was defined both for irrigated and rain-fed farmlands according to real information obtained from farmers and the Hamedan Agricultural Organization. Irrigation application is simulated based on an auto-irrigation routine, because it is difficult to know the volume and specific time of irrigation by farmers during simulation periods. Auto-irrigate triggers irrigation events based on water stress threshold. To estimate PET the Hargreaves method (Hargreaves & Samani 1985) is used. AET is obtained based on the Ritchie methodology (Ritchie 1972).

Several researchers have reported an uncertainty quantification of the hydrological models (Jin et al. 2010; Li et al. 2010, 2013). Li & Xu (2014) used Bayesian and generalized likelihood uncertainty estimation methods to estimate the uncertainty of a hydrological model. Abbaspour (2007) developed a SUFI-2 (sequential uncertainty fitting) method to analyse the uncertainty of the hydrological models.

After setting up a SWAT model run, a SUFI-2 algorithm is used for uncertainty analysis and to calibrate the model based on the monthly river discharge and yearly crop yield. The SUFI-2 represents all modeling sources of uncertainties such as input data, conceptual model and parameter selections. The uncertainties are mapped based on the parameter ranges. The algorithm tries to bracket most of the measured data within the 95% prediction uncertainty band (95PPU) which is calculated at 2.5 and 97.5% of the cumulative distribution of an output variable obtained through Latin hypercube sampling. To quantify the goodness of calibration and uncertainty performance two indices (P-factor and R-factor) are used. The P-factor is the percentage of data bracketed by the 95PPU band and the R-factor is the average width of the 95PPU band divided by the standard deviation (SD) of the measured data. To compare the measured and simulated monthly discharge and annual crop yield, Nash–Sutcliffe (NS) and root mean square error (RMSE) objective functions are used, respectively. In this research, we used a previously calibrated SWAT model of the Razan–Ghahavand river system. A detailed description about calibration, validation, uncertainty and sensitivity analysis is presented by Rafiei Emam et al. (2015a).

An increasing CO₂ concentration will have a severe effect on vegetation, especially on leaf area index development and stomata conductance (Wand et al. 1999) and, additionally, it can lead to changes in ET resulting in change to other water components.

The impact of climate change on water resources and crop production was assessed by developing a set of GCMs parameterized based on three IPCC-AR4 emission scenarios (A1B, B1 and A2) in the middle of the 21st century (2046–2065) defined by the Intergovernmental Panel on Climate Change (IPCC 2007a). The approved new set of scenarios is described in the IPCC Special Report on Emission Scenarios (SRES). Four different narrative storylines were developed to consistently describe the relationships between the forces driving emissions and their evolution and to add context to the scenario quantification. The resulting scenarios cover a wide range of the main demographic, economic and technological driving forces of future greenhouse gas and sulfur emissions. The B1 storyline and scenario family describes a convergent world with the same global population, that peaks in mid-century and declines thereafter, as in the A1 storyline, but with rapid change in economic structures toward a service and information economy, with reductions in material intensity and the introduction of clean and resource-efficient technologies. The emphasis is on global solutions to economic, social and environmental sustainability, including improved equity, but without additional climate initiatives. The three A1 groups are distinguished by their technological emphasis: fossil intensive (A1FI), non-fossil energy sources (A1 T), or a balance across all sources (A1B) (where balance is defined as not relying too heavily on one particular energy source, on the assumption that similar improvement rates apply to all energy supply and end-use technologies). The A2 storyline and scenario family describes a very heterogeneous world. The underlying theme is self-reliance and preservation of local identities. Fertility patterns across regions converge very slowly, which results in a continuously increasing population. Economic development is primarily regionally oriented and per capita economic growth and technologies change in a more fragmented and slower way than in other storylines.

Spatial resolution of GCM output is coarse (i.e., low resolution) therefore we applied a downscaling program (LARS-WG) developed by Semenov & Stratonovitch (2010) in order to generate fine resolution climate data to use the output in the hydrological model. LARS-WG is a model simulating time-series of daily weather at a single site. It utilizes semi-empirical distribution to obtain statistical parameters such as length of wet and dry periods, daily precipitation, daily minimum and maximum temperature and daily solar radiation based on daily observed climate data at local stations of the catchment. For the minimum and maximum temperature, auto- and cross-correlation calculated monthly were used by semi-empirical distribution calculation. For solar radiation, semi-empirical distributions with equal interval sizes are used. LARS-WG has been successfully tested in several case studies with diverse climate types (e.g., Lawless & Semenov 2005; Zarghami et al. 2011; Sunyer et al. 2012). The new version of LARS-WG includes the data for different GCM models. A number of statistical tests such as t-test, F-test and chi-squared test were implemented by LARS-WG in order to verify the results of the simulation by comparing the synthetic data with the observed data.

Our climate model provided downscaled output from four GCMs (CCSM3, CSIRO-MK3, MPEH5 and HADGEM). We developed a multi-model ensemble from these GCMs. An ensemble of climate models exhibits more reliable representation of regional and local uncertainties than results from individual GCMs by decreasing the biases from single models (Abbaspour et al. 2009; Gaiser et al. 2011; Wu et al. 2012). After downscaling GCM data, the daily climate data were put into the SWAT calibrated eco-hydrologic model to simulate the future impacts on water components. Parameter ranges in the hydrologic model represent the uncertainty of the model run. According to the calibrated model, 69 sets of eco-hydrologic parameters were used to capture the uncertainty of the model (e.g., CN2, alpha_bf). The model was then run for a baseline scenario (1998–2008) and future climate scenarios (2046–2065). The CO₂ concentration was defined as 492, 541 and 545 ppm for the selected climate scenarios B1, A1B and A2, respectively (Semenov & Stratonovitch 2010).

We estimated the impact of urbanization on surface runoff in our study area. To examine the land use change, two series of satellite images in 1989 and 2009 were used. The land use of 2009 was adopted for the SWAT model. Based on changes in land use during 20 years (1989–2009), future land use (2050) was predicted. Moreover, population and its growth rates were used as a driver in land use prediction (2046–2065).

As precipitation is already low in the southern lowlands of the watershed, the decline of precipitation may have a significant influence on increasing drought in the future. Hence, crop production and rangeland productivity in this area will be at a lower level which may lead to more land degradation and desertification. Rangelands in the northern and southern parts of the watershed are directly affected by precipitation and soil moisture distribution. The rangelands in the southern portion of the area are more sensitive than in the northern part due to lower precipitation and hence lower soil moisture.

Figure 6(e)–6(h) show historic patterns of the average maximum temperature (Figure 6(e)) and anomalies of the absolute difference between the maximum temperature prediction of the three scenarios and the average over the future (2046–2065) and historic (1998–2008) period. The results show the changes of maximum temperature in A1B and A2 are more than the B1 emission scenario. In scenarios A1B and A2, most of the north and central watershed experience an increase of about 2 °C in temperature (in scenario B1 about 1 °C) while in the south of the watershed the increase is mostly about 2.5 °C (in B1 scenario about 1.5 °C).

The diurnal temperature range is calculated as the difference between maximum and minimum temperature. For the historic period it is between 14 and 16.5 °C over the watershed area. This difference in the southern area is more than the central and the northern basin. However, the diurnal temperature will decrease in the north and central part of the watershed while it will increase in the south of the watershed during 2046–2065. Hence, the minimum temperature increase is stronger in the north and central part of watershed and lower in most parts of the southern water catchment.

ET varies considerably across the area due to soil and land cover variations. The annual actual ET varies from 233 to 415 mm with the highest ET in the northern part and lowest in the southern part of the watershed (Figure 6(i)). Approximately 294 mm of the water budget is lost by annual ET, which reveals that ET has the largest portion of the watershed's overall water budget. The percentage difference calculated between ET based on the emission scenarios and the baseline scenario shows that in most parts of the watershed the ET will decrease by up to 17%. However, an increase of ET is seen in the west of the watershed throughout all scenarios (Figure 6(j) and 6(k)). The B1 scenario indicated an increase in ET in the north and central part of the watershed while scenario A1B illustrated a further decrease in ET.

Figure 8(e)–8(h) show the annual spatial pattern of surface runoff in the study area. The northeast and the southwest of the watershed have significantly higher surface runoff due to the existence of low permeability soils. In scenario B1, the surface runoff rate is much higher than in other scenarios especially in the north, center and southwest of the watershed; the increase of surface runoff rises up to 65% in these regions. In the southeast, and some parts of the north of the watershed, the surface runoff decreased to 56%. However, in mountainous areas of the northern portion of the basin the runoff will increase.

The historical data show a soil moisture rate varying from 10 mm in the southwest to 80 mm in the northern portion of the watershed. According to the emission scenarios, soil moisture in the northern and southern portions of the watershed has significantly decreased whereas the central and western part of the area showed an increase of water content. The A1B scenario shows the highest decrease in soil moisture compared to the other scenarios (Figure 8(i)–8(l)).

Land use change monitoring indicates that urban or built-up land expanded by about 100% in Razan–Ghahavand during the 1989–2009 periods. The urban land was increased mostly due to the contraction of croplands. Based on this information and according to the mean annual population growth rate in the area reported by the Statistical Center of Iran (2011) of 2.63% during 1986–2006, we assumed the urban or built-up growth is about 190% during the 2046–2065 period. The results indicated that the average annual runoff volume increased by more than 60% from 2045 to 2064 due to urbanization. The increase of surface runoff by urbanization has already been mentioned in other studies (Weng 2001).

There are various land degradation phenomena in the study area. The evidence of these land degradations are salinization in the east and center, rangeland degradation in the south and subsidence in the west of the watershed. Land subsidence in the study area responds directly to groundwater resources. Owing to shortage of precipitation, groundwater is the main source of water resources for agriculture, industry and domestic purposes. In the west of the watershed due to the over-extraction of groundwater and sensitivity of the bedrock, land subsidences are occurring on a large scale.

Table 2 shows the estimation of average annual recharge in the historic period (1998–2008) and the percentage difference between the future groundwater recharge (2046–2065) and the baseline groundwater recharge. The highest decrease (24%) is based on the A1B scenario and the lowest decline (8%) is found for scenario B1. The field observations show that most land subsidence is happening in HRU 515 and 571 in the west of the basin. Therefore, the groundwater recharge was estimated in these areas specifically. The results show an increase of recharge in those places up to 52% for scenario B1. The reason for a rising recharge is due to increasing precipitation in this part of the basin. Overall, we can conclude that for the future the whole basin is facing both declining groundwater table and decreasing groundwater recharge.

This study investigated the ensemble climate change effect on hydrology, crop production and land degradation based on four GCMs in three emission scenarios. In addition, the impact of land use changes (urbanization) on runoff production was evaluated.

The Razan–Ghahavand basin is currently under high pressure due to growing water demand for drinking, agriculture and industrial purposes. The population has grown rapidly in the last 20 years. A future climate is likely to affect the water resources differently in the north and south of the basin. The spatial pattern of climate change models shows that with increasing CO₂ the temperature is increased and it leads to more ET and less groundwater recharge. A similar result is reported by Bouraoui et al. (2004). We found that precipitation in the south and central part of the basin will decline while the northern and the western parts will face more precipitation. The increase of precipitation in the west is the reason for rising groundwater recharge in HRUs with land subsidence phenomena; therefore, the water table in this part of the basin will increase. In other words, with good water management in this area further land subsidence could be controlled. However, the rangelands in the southern part of the basin are under water stress which will continue even in the future due to the decrease in precipitation. However, we can conclude that the watershed will have less rain and high summer temperatures. Therefore, a sustainable strategy needs to be considered for future water resources development.

On the other hand, wheat is the representative crop in the basin and suffered from heat stress during the historical period. Therefore, with rising temperatures in the future, the production will increase especially in irrigated lands. Nevertheless, production in rain-fed areas will decrease due to water stress. The water stress in rain-fed areas is greater than in irrigated areas; however, the uncertainty in irrigated lands is due more to the unaccounted water use in the model due to lack of data. Hence, the increase of water use efficiency will be mandatory in future cropping systems for the sustainable use of water resources.

Urbanization growth has a significant effect on surface runoff. It is important to note that to develop a water resources management plan, certain LULC changes (e.g., urbanization) as well as associated flood patterns should be considered.

Finally, it is important to note that the predictions of hydrological components in the future were based on the use of the same soil parameters for baseline. In this study, the change of soil parameters was not considered in the future. Change of soil parameters may have an effect on soil water content as well as surface runoff in the basin. Therefore, further research on climate change assessment while considering soil parameter changes would raise confidence on the outcomes. The same limitations have already been discussed by Faramarzi et al. (2013).

We used a set of four GCMs in this study; however, considering the fact that the effect of other GCMs or even RCMs may have different outcomes, they should be studied in further research.

The authors would like to express their gratitude to Dr Karim C. Abbaspour and Dr Samira Akhavan for their helpful comments during the calibration processes of the SWAT model. We are grateful to the editor and anonymous reviewers whose valuable comments and suggestions improved the manuscript greatly.