Climate change impact assessment on water resources in Iran: applying dynamic and statistical downscaling methods

In the contemporary era, study of the impact of climate variability and change, at different time and space scales, on earth's natural system and humanity has become the outstanding global priority of the international community. It is expected that almost every region in the world would experience a negative effect from climate change on water resources (IPCC 2007). In fact, there are variations in features and the intensity of the impact from one region to another. Some regions, mostly arid and semi-arid areas probably, would face water shortages. In addition, increasing demand due to surges in the number of people would increase the risk of water scarcity. Rising sea levels, increasing frequency of floods and droughts, high economic cost, a decrease in overall crop yield, an increase in poverty and hunger risk can be considered as effects of climate change in different parts of the world.

A significant number of publications deal with hydrological components, for instance, groundwater recharge (Jyrkama & Sykes 2007), streamflow (Caballero et al. 2007; Fu et al. 2007), runoff (Nunes et al. 2009) and evapotranspiration (Calanca et al. 2006). However, fewer publications have concentrated on the long-term climate change impact assessment on water resources by applying different downscaling approaches: statistical, the Long Ashton Research Station-Weather Generator (LARS-WG), the Model for the Assessment of Greenhouse-gas Induced Climate Change-SCENario GENerator (MAGICC-SCENGEN), and dynamic (Providing Regional Climates for Impacts Studies (PRECIS)). Furthermore, by applying downscaling approaches, an appropriate evaluation of climate change and water resources can be achieved that can be considered to be a beneficial application to support water resource planning and management (Serrat-Capdevila et al. 2007).

In this study, we present the climate change effect on water resources until 2100. Iran is highly vulnerable to the adverse impacts of climate change. Most parts of the country are arid or semi-arid, prone to drought and desertification, the limited forests are liable to decay, water resources are scarce, sea level rise is a threat to the very long coastal zones of the country, many urban and industrial areas are heavily polluted, and the country is mountainous with very fragile ecosystems. As a prerequisite to carrying out the vulnerability and adaptation (V&A) assessment, a study on climate variability and climate modeling was undertaken to predict the future climate of Iran based on the historical record. The Runoff Assessment Model (RAM) (Fahmi 1990) was applied to study the effect of climate change on runoff at a sub-basin level for the whole country.

Global climate model output does not have sufficient temporal and spatial resolution to be applied directly in hydrological models. Some approaches can be used to simulate daily climate change (Semenov & Barrow 1997). An inexpensive tool to downscale the climatic variables is LARS-WG (Semenov et al. 1998). The weather generator applies semi-empirical distributions, which presents greater flexibility in different climate reproductions. A study proved that LARS-WG acts well in various climates all around the world (Semenov et al. 1998).

Apart from LARS-WG, MAGICC-SCENGEN has been used to project future changes. MAGICC/SCENGEN is a coupled gas-cycle/climate model (MAGICC) that drives a spatial climate-change SCENGEN (Wigley 2008). MAGICC has been applied by IPCC in various assessments. There are possibilities to update and edit scenarios in models. Scenarios present data for general circulation atmospheric models and climatic global observation data for Asia, Europe and America (Abbasi et al. 2011). The SCENGEN model has been developed over several years. Greenhouse gas scenarios can be changed by SCENGEN and MAGICC in order to change the sea surface temperature and ground temperature (Abbasi et al. 2011).

Dynamic downscaling is another approach applied for producing climatic variables with higher resolution. PRECIS can be applied for the globe to produce detailed scenarios of climate change. PRECIS has been developed at the Hadley Centre, with the aim of introducing a flexible regional modeling system based on increasing demand for regional-scale climate projections (ECLAC 2010). PRECIS is based on the third generation of the Hadley Centre's regional climate model (Jones et al. 2004).

Generally, the aim of this study was threefold: first, to expand the knowledge of Iran's climate variability; second, to extend the knowledge beyond what has been achieved during the previous studies for the Initial National Communication (INC 2003) about the possible changes in the country's climate due to human interference with the earth's climate system; and, finally, to provide a V&A assessment of water resources with the necessary inputs. To achieve these defined goals and objectives, the study has been divided into four main sections: climate variability, climate change projection, downscaling and water resource impact assessment.

The Islamic Republic of Iran, with an area of 1,648,195 square km, is located in the southern part of the north moderate zone between 25 ° 03′ and 39 ° 47′ north latitude from the equator and 44 ° 05′ and 63.5 ° 18′ east longitude from the Greenwich meridian. Generally, Iran is a mountainous and semi-arid land, with a mean altitude of more than 1,200 meters above sea level.

The climate of the country is in the main influenced by a subtropical high-pressure belt. January (monthly average temperature in the range of −6 to 21 °C) and July (monthly average temperature in the range of 19 to 39 °C) are the coldest and the warmest months, respectively, in most cities of the country. Precipitation varies greatly nationwide and changes from season to season and year to year. Most climatic regions have their highest seasonal precipitation in winter. Except for the northwest, southeast, and Caspian Sea coasts, Iranian summers see no rain. The average annual total precipitation ranges between 50 mm to about 2,000 mm. According to the Koppen climate classification system, Iran can be divided into three climate types out of a total five types in its scheme. The dominant climate, covering 81% of the country, is of the arid and semi-arid subtype climates (B, dry main type). Different subtype climates of C (temperate-mesothermal main type) are experienced along the Caspian Sea coasts, and some places in the Zagros Mountains region located in the west of the country. Different subtypes of C cover 17% of the country. The rest of the country (2% coverage), which includes some small areas in the Alborz Mountains in the north and the Zagros Mountains in the west, is primarily climate type D (continental-microclimate main type).

Internal renewable water resources are estimated at 128.5 bcm/year. Surface runoff represents a total of 97.3 bcm/year and groundwater re-supply is estimated at about 49.3 bcm/year. Iran receives 6.7 bcm/year of surface water from Pakistan and 1.63 bcm/year from Azerbaijan, and an unquantified amount of water from Afghanistan via the Helmand River. The surface runoff to the sea and to other countries is estimated at 55.9 bcm/year. The total safe yield of groundwater (including non-renewable water or unknown groundwater inflows from other countries) has been estimated at 49.3 bcm/year. According to information related to forecasting national water consumption of the country, the annual renewable water availability per capita will be about 1,300 cubic meters in the year 2021, which, based on international measures, is considered crisis level. Also, due to population growth and reduced water resources, the per capita water availability has sharply decreased in the past half a century and will further decrease through 2021 (SNC 2010).

In this paper, climate change is assessed by means of two LARS-WG weather generators and all outputs from the available general circulation models (GCMs) in MAGICC-SCENGEN software in combination with different emission scenarios at the regional scale, while the PRECIS model has been used for projection at the local scale. Future climate scenarios for periods of 2010–2039 were generated from the ECHO-G model for scenario A1 by LARS-WG. Also, by MAGICC-SCENGEN, other future climate scenarios were simulated from HadCM2 and ECHAM4 in combination with 18 available emissions scenarios until the year 2100. As for the dynamic downscaling approach, by applying the PRECIS model, the time period of 2071–2100 was considered and future climate was generated for the period for A2 and B2 emission scenarios. A hydrological model, RAM, was first utilized to simulate water resources for Iran. Then, using the MAGICC-SCENGEN model and the downscaled results as input for the RAM model, a prediction was made for changes in 30 basins, and runoff.

LARS-WG is a stochastic weather generator which, under current and future climate conditions, can be applied for weather data simulation at a site (Racsko et al. 1991; Semenov et al. 1998). These data are daily time-series of climate variables, namely maximum and minimum temperature, precipitation and solar radiation. Observed daily weather for a specific site is applied in the weather generator as input to determine probability distribution parameters and correlations related to weather variables. The procedure of generation for synthetic weather data production is based on considering values from the suitable distributions by applying a pseudo-random number generator. Depending on whether the precipitation is greater than zero, dry and wet days can be distinguished by the weather generator. Precipitation is simulated by using semi-empirical monthly probability distributions for series of dry and wet days and for the precipitation amount on a wet day.

Cloud cover has a key role in minimum and maximum temperature and radiation simulations, and so wet and dry day distributions are used by LARS-WG for each of the aforementioned variables. For temperature variables, normal distribution is used with the standard deviation and mean varying daily based on finite Fourier series of order 3. Time auto-correlations applied for maximum and minimum temperature for the specific site are constant, and the standardized residuals cross-correlation from the daily mean is considered for all sites at 0.6. As for solar radiation, semi-empirical distributions with the same interval size are utilized (Semenov et al. 1998).

For a specific parameter set, the weather generator produces synthetic data for one day by determining the status of precipitation on that day. The data include intermittent dry and wet series, and at the end of one series the length of the next is selected from the semi-empirical distribution of dry or wet series for that month. The precipitation amount for a wet day is taken from the distribution of precipitation for the month, and the radiation and temperature values are taken from the distributions of wet days and applied correlation coefficients. On dry days, the distributions of dry day are used. By choosing one of the intervals and a value within that interval from the uniform distribution, values from the semi-empirical distributions are selected.

By applying semi-empirical distributions, flexibility is provided to the generator, letting it simulate a broad variety of distributions. LARS-WG has been examined for different places in the world including the USA, Asia and Europe. Results prove the model's ability to reproduce most of the observed data characteristics well (Semenov et al. 1998).

MAGICC/SCENGEN is a coupled gas-cycle/climate model (MAGICC) that drives a spatial climate-change SCENGEN (Wigley 2008). MAGICC can be considered as one of the primary models applied by IPCC since 1990 in order to obtain future sea level rise and global-mean temperature projections. The climate model in MAGICC is an energy-balance model, upwelling-diffusion, that produces hemispheric- and global-mean temperature and oceanic thermal expansion results. The MAGICC climate model is combined interactively with gas-cycle models that present projections related to key greenhouse gases concentrations.

In order to drive SCENGEN, global-mean temperatures from MAGICC are applied. A version of the pattern scaling method (Santer et al. 1990) is used by SCENGEN in order to simulate change spatial patterns from a data base of atmosphere/ocean GCM (AOGCM) data from the CMIP3/AR4 archive. The base of the pattern scaling method is the global-mean and future climate change spatial-pattern components separation, and the further separation of the aerosol components and the latter into greenhouse gas. Spatial patterns are normalized and expressed as variations per 1 °C shift in global-mean temperature. These normalized aerosol and greenhouse-gas components are suitably weighted, stepped up, and scaled up to the global-mean temperature that is described by MAGICC for a specific year, set of parameters of climate model, and emissions scenario. In the case of the SCENGEN scaling component, the user can choose from a number of AOGCMs for greenhouse-gas-induced climate patterns (Wigley 2008).

The PRECIS is a land surface and atmospheric model with high resolution that can be used over the globe. Determined surface and lateral boundary conditions are required by the model. Dynamic atmospheric data at the longitudinal and latitudinal edges of the model domain are provided by lateral boundary conditions, while conditions of the surface boundary are only required over water, on condition that the model requires time series of ice extents and surface temperatures. The conditions of lateral boundary include the components of horizontal wind, standard atmospheric variables of surface pressure and atmospheric temperature, and humidity measures. In addition, a boundary conditions set, including sulphate aerosols, sulphur dioxide and associated chemical species, is also needed, due to the fact that PRECIS contains a representation of the sulphur cycle. The model is defined in three major sections: (1) the dynamics; (2) the sulphur cycle; and (3) the physical parameterizations. The dynamics are related to the meteorological state variables advection modified by physical parameterizations: clouds, radiation, precipitation, boundary layer, gravity wave drag and surface exchanges. In addition, the sulphur cycle is a physical parameterization, but state variables of this cycle are considered as prognostic and advected as tracers. PRECIS is based on the third generation of the Hadley Centre's regional climate model. A complete description of PRECIS structure is presented by Jones et al. (2004).

The study reveals that the increase in minimum temperatures is more widespread than with the maximum temperatures. The discrepancies are remarkably higher in the large, heavily populated and industrialized cities. Due to the pattern of higher minimum temperatures, the daily temperature variability has reduced almost everywhere. The analysis of results shows that the temperature has risen between 2.5 and 5 °C during 1960–2005. There are also cities with clear temperature descent rates. Figures 6 and 7 illustrate the changes in the amount of precipitation in selected stations. Accordingly, it could be concluded that the southwestern part of the Caspian Sea, northwest and west of the country have experienced the highest rate of reduction in the amount of their annual precipitation. Study shows that the number of days with precipitation higher than 10 mm have reduced in the west, northwest, and southeast of the country. That number has increased in the other regions, except in the southeast of the Caspian Sea (Figure 7).

Figure 8 illustrates the changes in wind speed throughout the country between 1960 and 2005. The highest rates of decrease are seen in the central part of the country as well as the northeast. The highest increase is being seen in Zabol (southeast Iran).

Figure 9 illustrates humidity variability by considering the changes in dew point temperature. One of the most interesting results is related to the sharply ascendant dew point in the two cities of Bam and Shahr-é-Kord. Generally, it can be concluded that the dew point temperature is consistently lowering in most parts of the country. However, in the north and northeast the dew point average is clearly ascendant. The highest rise is seen in the city of Sabzevar. The dew point average rising in the north and northeast areas can be the result of an increase in temperature.

Figure 10 illustrates changes in daylight hours at selected stations. A rising rate pattern is visible everywhere throughout the country. The highest rate of increase is seen in the northwest of the country. Cloudiness is another important factor in the climate system. Figure 11 illustrates the trends in the days with clear skies. The results show that the number of days with clear skies changes between −12 to 12 per decade. The highest rise and fall in the number of days with clear skies are seen in a relatively small area of the country, in the cities of Shahroud and Gorgan, and could be due to the effect of the Alborz Mountains range on the climatic condition in different places.

The Climatological Research Institute had already conducted climate change studies at 43 synoptic stations throughout the country by means of LARS-WG (Semenov & Barrow 2002) in combination with the results of the A1 scenarios of the ECHO-G model (a GCM model that is being used in Hamburg University and the South Korea Center for Meteorological Research). This scenario has been used to project the country's climate during 2010–2039 and to compare the results with observations during 1976–2005. In the course of that study, changes in the number of dry, wet, freezing and hot days and extreme events like heavy and torrential rains, as well as changes in temperature and precipitation, have been examined.

To address uncertainties in climate change projections, the outcomes of different models and scenarios have been compared with each other. The results for the two HadCM2 and LARS-WG models with the two IS92a and A1 emission scenarios are presented here. To be able to compare the results of the projections, the baseline for climate change projection by MAGICC-SCENGEN was changed from 1961–1990 to 1976–2005. The results indicate that there are not any statistically meaningful differences between the projected changes in precipitation and temperature in the future according to those models and scenarios, at the 0.05 level of confidence. Table 1 shows the results of those studies for HadCM2 and LARS-WG.

PRECIS uses the output of GCMs as its boundary parameter to project future climate on a regional scale by producing higher resolution results (Jones et al. 2004). Therefore, it could be considered as a downscaling tool. Its outstanding difference from models like LARS-WG is that it uses the governing set of equations in the atmosphere to make the projection known as dynamical modeling.

The projection of precipitation using the A2 emission scenario suggests that the country will miss part of its available territorial water due to the reduction of precipitation (more specifically snowfall) over the west, southwest, and north along the Zagros and Alborz ranges along with an increase in runoff. Consequently, some of the important permanent rivers will face dryness as is already evident in the Zayandeh River. However, projection using the B2 emission scenario provides a different picture from changes in the amount of precipitation, but still shows a drastic reduction in snowfall over the northwest, west, southwest, and north of the country plus an increase in runoff that may be interpreted as a possible greater risk of flooding. Both scenarios project less summertime precipitation along the south coast of the Caspian Sea, which may have a great impact on the production of one of the very important crop yields of the region, i.e. rice.

In order to study the climate system and climate change on a global scale, GCMs are used. Results of GCM models show that this model, despite its limitations, including a lack of hydrological parameters for a simplified model, the heterogeneous spatial distribution of network data that is used as the input model, and finally the large scale model results, will most likely be the touchstone of future climate simulations. Using the MAGICC-SCENGEN model and downscaling the results to fit Iran's geography, rainfall and temperature conditions has been modeled until 2100. The calculations were performed for the IPCC scenarios HadCM2-18 and the results of the model have been used to predict surface runoff.

Using the results of the GCM model as input data for the RAM model, a prediction is made for changes in 30 basins and runoff (Fahmi 1990). All of the defined scenarios of GCM are used. However, the results of more appropriate scenarios for Iran's condition (namely IS92A, EMCONSR, IS92AEXT, IS92B, IS92C, IS92D, IS92E, IS92F, K-1PC, K-1CIN1, K-NOMOR, SRESA1, SRESA2, SRESB1, SRESB2, WRE450, WRE550 and WRE650) are shown in Tables 2–5. According to the studies conducted, temperature and precipitation changes are in the range of ±6 °C and ±60%, respectively (Fahmi 1990). The results of the model record that the rise in temperature not only increased evaporation and caused decreased runoff, but also accelerated melting snow, which causes an increased rate of runoff in winter and a corresponding decrease in runoff in spring. Model results also indicate that a constant level of rainfall of only about 2 degrees creates a rise of 27.3 bcm in annual volumes of evaporation and transpiration.

In general, based on the scenarios, Tshak-Bakhtegan and the Maharlu Basin see a probable 15.15% decrease in runoff while the Atrak Basin will witness a 7.24% increase. The greatest runoff decrease will be in the north provinces (Mazandaran and Qom), with 11.29%, and the highest increase will be in the northeast's North Khorasan Province at 6.33%.

Water supply is the crucial negative factor in national development given the drastic effects caused by the occurrence of flooding and drought. Thus, the impact of climate change on Iran's water sector is very important. The following actions can be effective in assessing the impact of climate change on water resources: enhance the monitoring networks of hydrological and meteorological data; comprehensive planning for optimized hydrological and meteorological monitoring networks is complete and recently put into implementation; application of simulation models with different scenarios that use GCM models to predict the future climate of Iran; and water and wastewater planning to provide safe water for arid and semi-arid areas.

Iran is highly vulnerable to the adverse impacts of climate change. It is a country with arid and semi-arid areas, limited water availability, low forest cover, liable to drought and desertification, prone to floods, high urban atmospheric pollution, fragile mountainous ecosystems and finally an economy highly dependent on production, processing and export of fossil fuels. This V&A study addressed climate variability and climate change modeling and impact on water resources.

Based on the meteorological data of 1960–2005, the minimum and maximum temperatures, precipitation (the amount and the number of days with precipitation higher than 10 mm), wind speed, dew point temperature (as an indicator of humidity), cloudiness and daylight hours have been studied in seasonal and annual timescales. The analyses of results demonstrate that temperature has risen between 2.5 and 5 °C on average, the increase in minimum temperatures is more widespread than the increase in maximum temperatures, with the discrepancies remarkably higher in large, heavily populated and industrialized cities, and, due to the pattern of higher minimum temperatures, the daily temperature variability has reduced almost everywhere. There are also cities with clear temperature descent rates. The southwestern parts of the Caspian Sea, northwest and west of the country have experienced the highest rate of reduction in the amount of their annual precipitation, i.e. the number of days with precipitation higher than 10 mm have reduced in the west, northwest, and southeast of the country, whereas precipitation has increased in other regions except in the southeast of the Caspian Sea. In addition, over the period 1960–2005, the highest rates of decrease in wind speed are seen in the central part of the country as well as in the northeast.

The dew point temperature, an indicator of humidity, has consistently decreased in most parts of the country except in the north and northeast parts. Furthermore, a rising rate pattern of daylight hours is visible everywhere throughout the country. The highest rate of increase is seen in the northwest of the country. As for cloudiness, the number of days with clear skies changes between −12 to 12 per decade, with the highest rise and fall in the number of days with clear skies observed in a relatively small area of the country in the cities of Shahroud and Gorgan, which could be due to the effect of the Alborz Mountains range on the climatic condition in different places.

According to the MAGICC-SCENGEN (HadCM2 and ECHAM4 Models in combination with 18 available emission scenarios), until the year 2100, both models predict a higher temperature nationwide with very little variation. The temperature will rise 0.4–3 °C based on one model and 0.5–4 °C based on another model. However, there are remarkable differences between the projected changes in precipitation and its spatial distribution. According to HadCM2, the northern half of the country will see a rise in the amount of precipitation, while the southern half of the country will suffer a net loss in precipitation.

According to the LARS-WG model, and based on the data of 43 synoptic weather stations, the climate of the country has also been forecast during 2010–2039 and the results have been compared with observations during the 1976–2005 period. The results indicate that the amount of precipitation will, on average, decrease throughout the country by 9% between 2010–2039 compared with the 1976–2005 period. However, the number of heavy and torrential rains will increase by 13% and 39% over the same period, respectively. Temperature projections show an average increase of 0.9 °C, and minimum and maximum temperatures will on average rise by 0.5 °C. The rise is more pronounced during the cold season. The number of hot days in most parts of Iran will increase. The highest increase will occur in the southeast of the country by 44.2 days. The study has also revealed that the number of freezing days in most parts of the country will decrease. The highest decrease will occur in the northwest of the country, with freezing days decreasing by 23 per annum. Study of the changes in the number of wet days during 2010–2039 indicates that it will increase in some areas in the northwest, center, south, east, and southeast of the country. In other parts of the country, the number of wet days will decrease. The highest decrease will occur in the cold season. The study of the number of dry days shows an increase in many parts of the country. The highest rise, at 36 days, is expected to occur in the west and southeast of the country.

The results of dynamic downscaling, by applying the PRECIS model, show the projected temperature changes using the A2 scenario to be on average higher by 2 °C than those of the B2 scenario. The precipitation projection using the A2 scenario shows that due to the reduction of precipitation, Iran will miss part of its available territorial water. However, the B2 emission scenario projection depicts a different view from variations in the amount of precipitation, but still shows a sharp reduction in snowfall over the northwest, west, southwest, and north of the country plus a rise in runoff that may result in greater flooding risk.

The RAM model results show that increasing temperature caused increasing evapotranspiration and decreasing runoff. Furthermore, increasing temperature caused a shorter snow melt period, which resulted in a runoff increase in winter and decrease in spring. At constant rainfall, the annual evapotranspiration volume increased by about 27.3 bcm with an increased temperature of only two degrees. Furthermore, the results prognosticate that except for three provinces, runoff in most basins will decrease. As an adaptation measure, and in order to preserve the already depleted groundwater resources and meet the increasing demand for water, it is planned to increase the utilization of surface water resources from the present level of 46 to 55% within the next 20 years. The hydroelectricity generation potential is over 25,000 MW, of which 6,700 MW is currently harvested with about 6,000 MW under construction, with a regulatory capacity of water dams with hydroelectric potential of about 41 million cubic meters in 2008. Due to the reduction in river runoffs, the efficiency of the hydropower plants will decrease, with adverse impacts on dam construction plans. As an another adaptation measure, construction of artificial aquifers using underground dams as a means of storing large quantities of good quality water is being considered.