• Al-Qinna, M., , Hammouri N. , , Obeidat M. , , and Ahmad F. , 2011: Drought analysis in Jordan under current and future climates. Climatic Change, 106, 421440, doi:10.1007/s10584-010-9954-y.

    • Search Google Scholar
    • Export Citation
  • Barnett, T. P., , Adam J. C. , , and Lettenmaier D. P. , 2005: Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438, 303309, doi:10.1038/nature04141.

    • Search Google Scholar
    • Export Citation
  • Caldwell, P., , Chin H.-N. , , Bader D. , , and Bala G. , 2009: Evaluation of a WRF dynamical downscaling simulation over California. Climatic Change, 95, 499521, doi:10.1007/s10584-009-9583-5.

    • Search Google Scholar
    • Export Citation
  • Christensen, J., , Kjellström E. , , Giorgi F. , , Lenderink G. , , and Rummukainen M. , 2010: Weight assignment in regional climate models. Climate Res., 44, 179194.

    • Search Google Scholar
    • Export Citation
  • Déqué, M., 2007: Frequency of precipitation and temperature extremes over France in an anthropogenic scenario: Model results and statistical correction according to observed values. Global Planet. Change, 57, 1626, doi:10.1016/j.gloplacha.2006.11.030.

    • Search Google Scholar
    • Export Citation
  • Déqué, M., 2009: Temperature and precipitation probability density functions in ENSEMBLES regional scenarios. ENSEMBLES Tech. Rep. 5, 63 pp.

  • Déqué, M., , and Somot S. , 2010: Weighted frequency distributions express modelling uncertainties in the ENSEMBLES regional climate experiments. Climate Res., 44, 195209, doi:10.3354/cr00866.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCAR mesoscale model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121, 14931513.

    • Search Google Scholar
    • Export Citation
  • Evans, J. P., 2009: 21st century climate change in the Middle East. Climatic Change, 92, 417432, doi:10.1007/s10584-008-9438-5.

  • Evans, J. P., , Smith R. B. , , and Oglesby R. J. , 2004: Middle East climate simulations and dominant precipitation processes. Int. J. Climatol., 24, 16711694, doi:10.1002/joc.1084.

    • Search Google Scholar
    • Export Citation
  • Fleury, P., , Plagnes V. , , and Bakalowicz M. , 2007: Modelling of the functioning of karst aquifers with a reservoir model: Application to Fontaine de Vaucluse (south of France). J. Hydrol., 345, 3849, doi:10.1016/j.jhydrol.2007.07.014.

    • Search Google Scholar
    • Export Citation
  • Gao, X., , Pal J. S. , , and Giorgi F. , 2006: Projected changes in mean and extreme precipitation over the Mediterranean region from a high resolution double nested RCM simulation. Geophys. Res. Lett., 33, L03706, doi:10.1029/2005GL024954.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., , and Lionello P. , 2008: Climate change projections for the Mediterranean region. Global Planet. Change, 63, 90104, doi:10.1016/j.gloplacha.2007.09.005.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., , Bi X. , , and Pal J. S. , 2004: Mean, interannual variability and trends in a regional climate change experiment over Europe. I. Present-day climate (1961–1990). Climatic Change, 22, 733756, doi:10.1007/s00382-004-0409-x.

    • Search Google Scholar
    • Export Citation
  • Hansen, J., , Ruedy R. , , Sato M. , , and Lo K. , 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.

  • Haylock, M. R., , Hofstra N. , , Klein Tank A. M. G. , , Klok E. J. , , Jones P. D. , , and New M. , 2008: A European daily high-resolution gridded dataset of surface temperature and precipitation for 1950–2006. J. Geophys. Res., 113, D20119, doi:10.1029/2008JD010201.

    • Search Google Scholar
    • Export Citation
  • Heckl, A., 2011: Impact of climate change on the water availability in the near east and the upper Jordan River catchment. Ph.D. thesis, University of Augsburg, 181 pp. [Available online at http://opus.bibliothek.uni-augsburg.de/opus4/files/1603/dissertation_heckl.pdf.]

  • Hertig, E., , and Jacobeit J. , 2008: Downscaling future climate change: Temperature scenarios for the Mediterranean area. Global Planet. Change, 63, 127131.

    • Search Google Scholar
    • Export Citation
  • Hu, C., , Hao Y. , , Yeh T. , , Pang B. , , and Wu Z. , 2008: Simulation of spring flows from a karst aquifer with an artificial neural network. Hydrol. Processes, 22, 596604.

    • Search Google Scholar
    • Export Citation
  • Kafle, H., , and Bruins H. , 2009: Climatic trends in Israel 1970–2002: Warmer and increasing aridity inland. Climatic Change, 96, 6377, doi:10.1007/s10584-009-9578-2.

    • Search Google Scholar
    • Export Citation
  • Kattan, Z., 1997: Environmental isotope study of the major karst springs in Damascus limestone aquifer systems: Case of the Figeh and Barada springs. J. Hydrol., 193, 161182, doi:10.1016/S0022-1694(96)03137-X.

    • Search Google Scholar
    • Export Citation
  • Klein Tank, A., , Manzini E. , , Braconnot P. , , Doblas-Reyes F. , , Buishand T. , , and Morse A. , 2009: Evaluation of the ENSEMBLES prediction system. ENSEMBLES—Climate change and its impacts: Summary of research and results from the ENSEMBLES project, Final Rep., P. Van der Linden and J. Mitchell, Eds., Met Office Hadley Centre, Exeter, UK, 95–106. [Available online at http://ensembles-eu.metoffice.com/docs/Ensembles_final_report_Nov09.pdf.]

  • Knutti, R., , Furrer R. , , Tebaldi C. , , Cermak J. , , and Meehl G. A. , 2010: Challenges in combining projections from multiple climate models. J. Climate, 23, 27392758.

    • Search Google Scholar
    • Export Citation
  • Kunstmann, H., , Heckl A. , , and Rimmer A. , 2006: Physically based distributed hydrological modelling of the upper Jordan catchment and investigation of effective model equations. Adv. Geosci., 9, 123130.

    • Search Google Scholar
    • Export Citation
  • Kunstmann, H., , Suppan P. , , Heckl A. , , and Rimmer A. , 2007: Regional climate change in the Middle East and impact on hydrology in the upper Jordan catchment. IAHS Publ., 313, 141149.

    • Search Google Scholar
    • Export Citation
  • Kurtulus, B., , and Razack M. , 2010: Modeling daily discharge responses of a large karstic aquifer using soft computing methods: Artificial neural network and neuro-fuzzy. J. Hydrol., 381, 101111.

    • Search Google Scholar
    • Export Citation
  • LaMoreaux, P. E., , Hughes T. H. , , Memon B. A. , , and Lineback N. , 1989: Hydrogeologic assessment—Figeh spring, Damascus, Syria. Environ. Geol. Water Sci., 13, 73127, doi:10.1007/BF01664696.

    • Search Google Scholar
    • Export Citation
  • Lionello, P., , and Giorgi F. , 2007: Winter precipitation and cyclones in the Mediterranean region: Future climate scenarios in a regional simulation. Adv. Geosci., 12, 153158.

    • Search Google Scholar
    • Export Citation
  • Mariotti, A., , Zeng N. , , Yoon J. , , Artale V. , , Navarra A. , , Alpert P. , , and Li L. , 2008: Mediterranean water cycle changes: Transition to drier 21st century conditions in observations and CMIP3 simulation. Environ. Res. Lett., 3, 044001, doi:10.1088/1748-9326/3/4/044001.

    • Search Google Scholar
    • Export Citation
  • Meslmani, Y., 2009: Vulnerability assessment and adaptation measures to climate change in Syria. Final Rep., General Commission of Environmental Affairs, Damascus, Syria, 54 pp. [Available online at http://static.weadapt.org/placemarks/files/584/4dda6010f2e17INC-SY_V&A_General_Assessment_-En.pdf.]

  • Murphy, J. M., , Sexton D. M. H. , , Barnett D. N. , , Jones G. S. , , Webb M. J. , , Collins M. , , and Stainforth D. A. , 2004: Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature, 430, 768772, doi:10.1038/nature02771.

    • Search Google Scholar
    • Export Citation
  • Nakicenovic, N., , and Swart R. , Eds., 2000: Special Report on Emissions Scenarios. Cambridge University Press, 599 pp.

  • Ohmura, A., , and Wild M. , 2002: Is the hydrological cycle accelerating? Science, 298, 13451346, doi:10.1126/science.1078972.

  • Önol, B., , and Semazzi F. H. M. , 2009: Regionalization of climate change simulations over the eastern Mediterranean. J. Climate, 22, 19441961.

    • Search Google Scholar
    • Export Citation
  • Pöhler, H. A., 2006: Anpassung von WaSiM-ETH und die Erstellung und Berechnung von Landnutzungs- und Klimaszenarien für die Niederschlag-Abfluss-Modellierung am Biespiel des Osterzgebirges. Ph.D. thesis, Technische Universität Bergakademie Freiberg, 134 pp. [Available online at http://www.wasklim.de/download/Dissertation_Poehler.pdf.]

  • Rowell, D. P., 2005: A scenario of European climate change for the late twenty-first century: Seasonal means and interannual variability. Climate Dyn., 25, 837849, doi:10.1007/s00382-005-0068-6.

    • Search Google Scholar
    • Export Citation
  • Samuels, R., , Smiatek G. , , Krichak S. , , Kunstmann H. , , and Alpert P. , 2011: Extreme value indicators in highly resolved climate change simulations for the Jordan River area. J. Geophys. Res., 116, D24123, doi:10.1029/2011JD016322.

    • Search Google Scholar
    • Export Citation
  • Schulla, J., , and Jasper K. , 2007: Model description WaSiM-ETH. Tech. Rep., Institute for Geography, ETH Zürich, 181 pp.

  • Smadi, M., , and Zghoul A. , 2006: A sudden change in rainfall characteristics in Amman, Jordan during the mid 1950s. Amer. J. Environ. Sci., 2, 8491.

    • Search Google Scholar
    • Export Citation
  • Smiatek, G., , Kunstmann H. , , and Heckl A. , 2011: High-resolution climate change simulations for the Jordan River area. J. Geophys. Res., 116, D16111, doi:10.1029/2010JD015313.

    • Search Google Scholar
    • Export Citation
  • Soltani, S., , Saboohi R. , , and Yaghmaei L. , 2012: Rainfall and rainy days trend in Iran. Climatic Change, 110, 187213, doi:10.1007/s10584-011-0146-1.

    • Search Google Scholar
    • Export Citation
  • Somot, S., , Sevault F. , , Déqué M. , , and Crépon M. , 2008: 21st century climate change scenario for the Mediterranean using a coupled atmosphere–ocean regional climate model. Global Planet. Change, 63, 112126, doi:10.1016/j.gloplacha.2007.10.003.

    • Search Google Scholar
    • Export Citation
  • Sowers, J., , Vengosh A. , , and Weinthal E. , 2011: Climate change, water resources, and the politics of adaptation in the Middle East and North Africa. Climatic Change, 104, 599627, doi:10.1007/s10584-010-9835-4.

    • Search Google Scholar
    • Export Citation
  • Stott, P. A., , and Forest C. E. , 2007: Ensemble climate predictions using climate models and observational constraints. Philos. Trans. Roy. Soc. London, 365A, 20292052, doi:10.1098/rsta.2007.2075.

    • Search Google Scholar
    • Export Citation
  • Tebaldi, C., , and Knutti R. , 2007: The use of the multi-model ensemble in probabilistic climate projections. Philos. Trans. Roy. Soc. London, 365A, 20532075, doi:10.1098/rsta.2007.2076.

    • Search Google Scholar
    • Export Citation
  • Trichakis, I., , Nikolos I. , , and Karatzas G. , 2011: Artificial neural network (ANN) based modeling for karstic groundwater level simulation. Water Resour. Manage., 25, 11431152.

    • Search Google Scholar
    • Export Citation
  • Trigo, R. M., , Gouveia C. M. , , and Barriopedro D. , 2010: The intense 2007–2009 drought in the Fertile Crescent: Impacts and associated atmospheric circulation. Agric. For. Meteor., 150, 12451257, doi:10.1016/j.agrformet.2010.05.006.

    • Search Google Scholar
    • Export Citation
  • Weigel, A. P., , Knutti R. , , Liniger M. A. , , and Appenzeller C. , 2010: Risks of model weighting in multimodel climate projections. J. Climate, 23, 41754191.

    • Search Google Scholar
    • Export Citation
  • Xoplaki, E., , González-Rouco J. F. , , Luterbacher J. , , and Wanner H. , 2004: Wet season Mediterranean precipitation variability: Influence of large-scale dynamics and trends. Climate Dyn., 23, 6378, doi:10.1007/s00382-004-0422-0.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Orientation map of the Figeh area. FI is the extent of the investigation area used in the climate study and grid cells a and b are the areas used in the ANN. RCM grid points refer to the MM5 model runs in Lambert’s projection.

  • View in gallery

    Monthly statistical values of the Figeh spring flow intensity (1985–2006).

  • View in gallery

    Hydrological year (September–August) precipitation anomalies from the 1961–90 mean for five meteorological stations (with meters above mean sea level provided) and the average annual precipitation of the five stations (Figeh average). The thick gray line is the locally weighted scatterplot smoothing line.

  • View in gallery

    Cross correlation results for the 5-day time window and the 60-day time window used in assessment of the time-delay effects on the Figeh spring discharge: the letters a and b refer to the grid cells of the applied RCM data, as shown in Fig. 1.

  • View in gallery

    Simulated and observed Figeh discharge for (a) the calibration period and (b) the validation period.

  • View in gallery

    Observed (OBS) and interpolated (E-OBS) data from the E-OBS dataset and simulated monthly ensemble mean (ENS) precipitation in the Figeh spring area.

  • View in gallery

    (a) Multimodel temperature and (b) precipitation anomalies from the 1961–90 mean. The thin red and blue lines show the single-model realizations; the black line shows the multimodel 10-yr running average.

  • View in gallery

    Simulated monthly ensemble (a) mean temperature and (b) snow cover in the periods 1961–90 (circles), 2021–50 (triangles), and 2070–99 (squares).

  • View in gallery

    Measurements of P, E, and PE. Solid lines represent 1961–90, dashed lines represent 2021–50, and dotted lines represent 2070–99.

  • View in gallery

    Probability density functions of the response value R for precipitation in the Figeh area for (a) 2021–50 and (b) 2070–99 relative to 1961–90.

  • View in gallery

    Change of (a) MQ 2021–50, (b) MQ 2069–98, (c) MNQ 2021–50, and (d) MNQ 2069–98 relative to 1961–90. Model acronyms are explained in Table 3.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 110 110 13
PDF Downloads 88 88 5

Hydrological Climate Change Impact Analysis for the Figeh Spring near Damascus, Syria

View More View Less
  • 1 Atmospheric Environmental Research, Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology, Garmisch-Partenkirchen, Germany
  • | 2 Institute for Geography, University of Augsburg, Augsburg, Germany
  • | 3 Atmospheric Environmental Research, Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology, Garmisch-Partenkirchen, and Institute for Geography, University of Augsburg, Augsburg, Germany
© Get Permissions
Full access

Abstract

A set of downscaled climate change data from transient experiments with regional climate models has been used to access the future climate change signal in the area of the Figeh spring system in Syria and its potential effects on future water availability. The data ensemble at a spatial resolution of 0.25° has been investigated for the period 1961–90 for present-day climate and the periods 2021–50 and 2070–99 for future climate. The focus is on changes to annual, seasonal, and monthly surface air temperature and precipitation. For the first time, the Figeh spring discharge has been assessed with a hydrological runoff model based on an artificial neural network (ANN) approach. The ANN model was formulated and validated for the years 1987–2007, applying daily meteorological driving data. The investigations show that water supply from the spring might face serious problems under changed climate conditions. An expected, a precipitation decrease of about −11% in winter and −8% in spring, together with increased temperatures of up to +1.6°C and a significant decrease in snow mass, can substantially limit the water recharge potential already in the near future until 2050. In the period 2070–99, the annual precipitation amount is simulated to decrease by −22% and the annual mean temperature to increase by +4°C, relative to the 1961–90 mean. The ensemble mean of the relative change in mean discharge reveals a decrease during the peak flow from March to May, with values up to −20% in 2021–50 and almost −50% in the period 2069–98, both related to the 1961–90 mean.

Corresponding author address: Gerhard Smiatek, IMK-IFU, Karlsruhe Institute of Technology, Kreuzeckbahnstrasse 19, 82467 Garmisch-Partenkirchen, Germany. E-mail: gerhard.smiatek@kit.edu

Abstract

A set of downscaled climate change data from transient experiments with regional climate models has been used to access the future climate change signal in the area of the Figeh spring system in Syria and its potential effects on future water availability. The data ensemble at a spatial resolution of 0.25° has been investigated for the period 1961–90 for present-day climate and the periods 2021–50 and 2070–99 for future climate. The focus is on changes to annual, seasonal, and monthly surface air temperature and precipitation. For the first time, the Figeh spring discharge has been assessed with a hydrological runoff model based on an artificial neural network (ANN) approach. The ANN model was formulated and validated for the years 1987–2007, applying daily meteorological driving data. The investigations show that water supply from the spring might face serious problems under changed climate conditions. An expected, a precipitation decrease of about −11% in winter and −8% in spring, together with increased temperatures of up to +1.6°C and a significant decrease in snow mass, can substantially limit the water recharge potential already in the near future until 2050. In the period 2070–99, the annual precipitation amount is simulated to decrease by −22% and the annual mean temperature to increase by +4°C, relative to the 1961–90 mean. The ensemble mean of the relative change in mean discharge reveals a decrease during the peak flow from March to May, with values up to −20% in 2021–50 and almost −50% in the period 2069–98, both related to the 1961–90 mean.

Corresponding author address: Gerhard Smiatek, IMK-IFU, Karlsruhe Institute of Technology, Kreuzeckbahnstrasse 19, 82467 Garmisch-Partenkirchen, Germany. E-mail: gerhard.smiatek@kit.edu

1. Introduction

The Figeh spring, located 16 km northwest of Damascus, Syria, is one of the world’s largest springs and is the major water source serving a population of over 2.9 million spread over an area of 160 km2. Decreasing precipitation trends observed in various parts of the eastern Mediterranean region (EM), together with recent drought periods, increased concerns about water availability under changing future climate conditions. In 2010, a record global 12-month running mean temperature high was reached (Hansen et al. 2010), and the EM region was going through its worst rainy seasons since 1958. Intense droughts significantly affected the area of the Fertile Crescent in the previous three years, especially in the year 2008 (Trigo et al. 2010). Regional climate models (e.g., Evans et al. 2004; Giorgi et al. 2004; Lionello and Giorgi 2007; Kunstmann et al. 2007; Giorgi and Lionello 2008; Somot et al. 2008; Hertig and Jacobeit 2008; Önol and Semazzi 2009; Smiatek et al. 2011; Samuels et al. 2011) indicate further temperature increases and a precipitation decrease in the range of −20% for the EM.

In the present study, for the first time, a hydrological model able to describe the observed daily discharge by meteorological driving only was developed for the data-sparse and hydrogeologically complex Figeh spring. It is applied in combination with an ensemble of four dynamically downscaled and spatially highly resolved climate change datasets to analyze expected future climate change signals to the Figeh spring recharge area and their effect on the recharge cycle in an artificial neural network (ANN) approach. The single datasets result from RCM (regional climate model) experiments performed at the Karlsruhe Institute of Technology Institute of Meteorology and Climate Research–Atmospheric Environmental Research (KIT/IMK-IFU) within the frame of the Global Change and the Hydrological Cycle (GLOWA) Jordan River (JR) project (http://www.glowa-jordan-river.de). GLOWA JR is an interdisciplinary research project providing scientific support for sustainable water management in the Near East. Both climate change investigations, as well as the hydrological application, play an important role in support of local authorities in management and protection of the Figeh spring system. They can also support the extension of capacities for adaptive governance of water scarcity, which is presently underdeveloped in the area (Sowers et al. 2011).

Additional RCM data from the Ensemble-Based Predictions of Climate Changes and their Impacts (ENSEMBLES) Sixth Framework Programme (FP6) program (http://ensembles-eu.metoffice.com) were used to extend and strengthen the information on the simulated future climate change signal. ENSEMBLES FP6 provides ensemble-based simulations of future climate that were generated at various European institutions. In total, six different RCMs driven with lateral boundary data from three different global circulation models (GCMs) allowed for an extended view on the major component of the recharge cycle, such as precipitation amount and its interannual variability, evaporation, snow cover, and temperature.

Facing a highly complex karstic spring structure and a very limited hydrological and meteorological data availability, application of physically based hydrological models is not feasible. Therefore, the application focus was set on development of a hydrological model for the Figeh spring system that is based on an ANN architecture and daily time scale. The aim of this study was to find a suitable transfer function between the meteorological input data and the spring discharge and to apply it to the simulated climate change data from dynamical downscaling experiments, finally allowing a hydrological climate change impact assessment for the Figeh spring discharge.

In section 2, our study presents the Figeh area and the available regional climate change data, as well as the ANN architecture. Section 3 discusses the simulated expected changes in climate variables with a focus on the Damascus area and the Figeh spring discharge for two future periods, 2021–50 and 2070–99. Here, a multimodel ensemble approach and probability density functions (PDFs) are used to derive and analyze the possible changes in precipitation amount, variability, surface temperature, snow mass, and evaporation. The hydrological statistics include mean monthly flow (MQ), mean monthly low flow (MNQ), mean monthly high flows (MHQ), and annual discharge volumes. Conclusions are presented in section 4.

2. Material and methods

a. The Figeh spring system area

Three springs, the Figeh main spring, the Figeh side spring, and the Harouch spring, form the Figeh spring system of the karst Damascus limestone aquifer. Its groundwater bodies mainly originate from direct infiltration of atmospheric water, with a mean turnover time of 50 years (Kattan 1997). The recharge area of the Figeh spring system is located in the Anti-Lebanon range and comprises an estimated area of approximately 700 km2; however, its exact extent remains unknown. The elevations range from 1000 to over 2500 m MSL (see Fig. 1) and the climate is Mediterranean, with dry, hot summers and cool, wet winters and two transitional periods in autumn and spring. The geomorphology is the result of Jurassic-to-recent deposition, volcanism, and tectonics (LaMoreaux et al. 1989). Precipitation occurs predominantly in the months from October to April, with a peak in January. The annual mean precipitation in the mountainous region is in the range of 500 mm yr−1. The annual mean temperatures vary between 4.5° at the highest elevations and 17°C at 1000-m elevation (Kattan 1997).

Fig. 1.
Fig. 1.

Orientation map of the Figeh area. FI is the extent of the investigation area used in the climate study and grid cells a and b are the areas used in the ANN. RCM grid points refer to the MM5 model runs in Lambert’s projection.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

b. Observational reference data

Daily spring discharge data were obtained from records of the Damascus City Water Supply and Sewerage Authority (DAWSSA) for the episode from 1 March 1984 to 8 September 2009. The average Figeh flow intensity is 6.3 m3 s−1. Peak values of up to 25.9 m3 s−1 are typically monitored in April, and low flows occur in the summer season. In past decades, the spring system never ran dry but reached low flow values of only 1.4 m3 s−1. The numbers have to be seen in the context of water amounts abstracted from the spring system by associated wells with episodic pumping of 3.5–5 m3 s−1. There is a substantial interannual variability between 121 × 106 m3 yr−1 and 349 × 106 m3 yr−1 in Figeh discharge; MQ, MNQ, and MHQ are illustrated in Fig. 2.

Fig. 2.
Fig. 2.

Monthly statistical values of the Figeh spring flow intensity (1985–2006).

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Observational meteorological data reference for various periods from 1935 to 2009 results from station data provided by DAWSSA and the fifth version of the European daily high-resolution gridded dataset (E-OBS) (Haylock et al. 2008). There is only a limited number of stations with data available in the region, and thus, a limited representativeness for the complex terrain can be expected here. Five meteorological stations were available in the area of the spring (see Fig. 1). The stations Serghaya (elevation 1400 m MSL), Zabadani (1160 m MSL), Madaya (1105 m MSL), and Bloudan (1540 m MSL) are located west of the recharge area and Harireh (1520 m MSL) is located close to the southern outlet. Averaging of the data from the five available stations for the period 1956–2009 yields an average annual precipitation of 515 mm, with a minimum of 273 mm and maximum of 864 mm. The coefficient of variation (standard deviation divided by mean value) is 0.23. In the period 1961–90, the highest annual mean precipitation was measured at the Serghaya station with 543 mm yr−1, with 57% falling in the winter season [December–February (DJF)], 26% in spring [March–May (MAM)], and 16% in autumn [September–November (SON)]. The summer season [June–August (JJA)] contributed only 1% to the total precipitation. The last event with precipitation in the summer season was observed in the hydrological year 1991/1992. The highest precipitation since 1956 was observed in the hydrological year 2002/2003, with an amount of 1065 mm yr−1.

LaMoreaux et al. (1989) point out that 88% of the Figeh recharge area is at an elevation higher than 1500 m MSL and more than 60% is as high as 1900 m MSL. Therefore, the available monitoring network may not be representative in capturing the meteorological characteristics of the entire catchment. On the other hand, the true recharge area is still unknown, and present assumptions may substantially overestimate its extent. Stable isotope investigations performed by Kattan (1997) indicated that the recharge zones of the Figeh main spring are at elevations of around 1750 m MSL and higher but are lower for the side spring (1500 m MSL) and the Harouch spring (1300 m MSL). The same author points out the complexity of the hydrogeologic structure of the Figeh karst, which presumably consists of two different but interrelated flow systems. Numerous water losses, such as underflow to the Barada River and other areas, perennial springs, seeps, and withdrawal of groundwater from wells, as well as limited climatological data often not representative of the climate in the complex terrain with extremely sharp gradients, add to this complexity.

Xoplaki et al. (2004) found mixed rainfall trends for the EM, but clearly more stations showed decreasing trends in Greece, Turkey, Syria, and Lebanon for the period 1951–90. While Soltani et al. (2012) found no significant trends in precipitation over Iran, Kafle and Bruins (2009) conclude from climatic trends in Israel during the period 1970–2002 that the climate has become more arid in large parts of the country, with the exception of the coastal region.

Meslmani (2009) investigated recent temperature and precipitation trends in long-time records of 28 selected meteorological stations in Syria by applying the Mann–Kendall trend test. Coherent areas of significant precipitation decreases were found in winter and increases in autumn. The annual trends are rather insignificant, but the station majority shows precipitation decreases. Almost all stations show significant increases in the surface temperature in all seasons. In the period 1955–2006, monitoring stations located in the Figeh vicinity reveal insignificant mixed trends in precipitation and more significant trends in temperature.

In the Figeh area itself, the situation is similar (see Fig. 3). All stations depict a minor decreasing trend since the mid-1980s. Smadi and Zghoul (2006) report a sudden change and shift in the number of rainy days and the total precipitation in Amman (Jordan) during the mid-1950s. Significant precipitation decreases in this period are also visible at the Zabadani station, where records are available from 1935; until 1970, precipitation increases were observed.

Fig. 3.
Fig. 3.

Hydrological year (September–August) precipitation anomalies from the 1961–90 mean for five meteorological stations (with meters above mean sea level provided) and the average annual precipitation of the five stations (Figeh average). The thick gray line is the locally weighted scatterplot smoothing line.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Groundwater modeling has been applied to the adjacent Zabadani basin, covering an area of approximately 140 km2 (Meslmani 2009). Here, the impact of the projected climate change on groundwater level was simulated by the application of the modular three-dimensional finite-difference groundwater flow model (MODFLOW) linked to the Water Evaluation and Planning (WEAP). In addition, a hydrogeological conceptual model of the Barada spring, the largest spring in the Zabadani catchment, has been established. It is based on recharge of the karstic system by precipitation and snow melting and a two-discharge-reservoirs concept that was successfully applied to karst springs in Europe by Fleury et al. (2007). The study concludes a decreasing trend in the spring discharge and advocates further efforts in both improvements in the collection and dissemination of observational reference data, as well as the simulation of the hydrological cycle at scales appropriate to decision makers.

Special attention should be paid, however, to the application of regional climate modeling results as obtained, for example, by dynamic downscaling experiments of global climate models. Kunstmann et al. (2006) investigated in the upper Jordan area if observed and simulated runoff components can be explained by simple lumped approaches based on linear filter theory and neural networks. They state that the simulated runoff components can effectively be described by precipitation and a small number of adjustable parameters.

c. Applied hydrological model

Because of the complex hydrogeology of the Figeh spring system area and the very limited data availability, physical simulation of the spring discharge is not feasible. For that reason, a black box model approach based on an artificial neural network was developed. Black box models provide a mathematical function, transferring one or several input parameters into one or more output parameters (e.g., Pöhler 2006). Recent studies on karstic spring flow modeling were done, for example, by Hu et al. (2008) and Kurtulus and Razack (2010). Hu et al. (2008) simulated the Niangziguan karstic spring flow with ANNs in China using a multilayer feedforward neural network trained by a back-propagation algorithm driven with precipitation as input. Kurtulus and Razack (2010) simulated the La Rochefoucauld karst spring in the southwest of France on a daily time scale. Two types of soft computing models were used there to describe the spring behavior: an adaptive neural fuzzy inference system (ANFIS) based on fuzzy rules and an artificial neural network approach based on a feedforward back-propagation network trained with the Levenberg–Marquardt algorithm. Daily values of precipitation, piezometric level, and previous discharge were used as input. Some new approaches utilize direct human intervention as an input parameter. Trichakis et al. (2011) modeled the behavior of the karstic Edward’s aquifer in Texas (United States) with a multilayer perceptron using rainfall and temperature data but also with hydrogeological parameters like pumping rates from nearby wells.

With respect to the Figeh spring system, neural networks provide clear advantages: difficult processes, such as hydrological interactions in the karst system, do not have to be understood in detail; representation of the spring characteristics is possible with limited available data; and training and simulation are less time consuming than in physical models. The major disadvantage is the fact that the mathematical model equation cannot be interpreted physically; thus, it provides only limited system information.

The developed network structure consists of one input, one hidden, and one output layer. Six different input variables/neurons and nine hidden-layer neurons were chosen. A log-sigmoid transfer function is the hidden layer and a linear transfer function is the activation function of the output layer. Training was performed with the Levenberg–Marquardt algorithm. Two variables, precipitation and a substitute variable for real evapotranspiration called evapotranspiration index, were used as input into the ANN. The evapotranspiration index is calculated as the product of daily potential evapotranspiration, computed in dependence of the temperature-applying method provided by Hamon (e.g., Schulla and Jasper 2007), and precipitation. As in the karstic terrain of the Figeh spring catchment area, water infiltrates very fast (LaMoreaux et al. 1989), so evapotranspiration only needs to be considered in short time periods.

In preprocessing, time-delay effects on the spring discharge are taken into account, with a cross correlation of the input and output and an adequate modification of the input time series according to the delay with the highest correlation coefficient. An appropriate relationship between the precipitation events and Figeh discharge was derived by consideration of time windows. Here, the values are calculated daily as the sum of precipitation in a defined time window of the past.

Among various investigated choices, the following three different time windows for precipitation were identified as suitable: a 5-day time window for short-term discharge variability (see grid cells a and b in Fig. 1), a 60-day time window for midterm discharge variations (grid cells a and b), and a 365-day time window to reproduce the base flow (grid cell average). For evapotranspiration index, one 60-day time window for midterm discharge variations (grid cell average) was defined. The optimum time delays were derived by consideration of a time interval of 0–200 days. For each day, the precipitation dataset was shifted day by day backward in time and correlated with the discharge intensity. Figure 4 shows the results for the 5-day and 60-day windows. The ANN input time series are then shifted in relation to the delay with the highest correlation coefficient, that is, 97 days for the 5-day window and 72 (cell a) and 70 days (cell b) for the 60-day window. Finally, following Kurtulus and Razack (2010), all input variables and the Figeh spring discharge intensity were scaled into the range between 0.1 and 0.9.

Fig. 4.
Fig. 4.

Cross correlation results for the 5-day time window and the 60-day time window used in assessment of the time-delay effects on the Figeh spring discharge: the letters a and b refer to the grid cells of the applied RCM data, as shown in Fig. 1.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Figures 5a and 5b demonstrate that the developed ANN driven with daily meteorological input data is able to satisfactorily simulate the Figeh spring flow intensity. Small residual errors and comparatively high Nash–Sutcliffe (see Table 1) efficiency support the reliability of the ANN. Table 2 compares the observed and observational reference ANN-simulated discharge rates at the Figeh spring. The results illustrate the large interannual discharge variability and the reproduction of the annual discharge rates by the ANN between −23% and +14%. In the period 1988–2003, the average observed discharge was 207 × 106 m3, and the average simulated discharge underestimates it with 198.5 × 106 m3 by −4.1%. The explicit ANN formula is given in the appendix.

Fig. 5.
Fig. 5.

Simulated and observed Figeh discharge for (a) the calibration period and (b) the validation period.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Table 1.

Performance parameter of the different ANN parts: r is correlation, ME is mean residual error, MAE is mean absolute residual error, RMSE is root-mean-square residual error, and NS is Nash–Sutcliffe square error.

Table 1.
Table 2.

Observed and ANN-simulated annual discharge volumes at the Figeh spring. Anomaly is anomaly from the 1988–2003 mean, and difference is the simulated vs observed difference.

Table 2.

d. Investigated parameters

The investigated climatic parameters and statistics reflect the assessment of the future water recharge potential and include monthly and seasonal mean temperature, precipitation, snow cover, evaporation, and interannual precipitation variability represented as the coefficient of variation. The investigated period ranges from 1961 to 2099, with particular intervals at 1961–90 for the present and 2021–50 and 2070–99 for future climate. The Hadley Centre Coupled Model, version 3 (HadCM3), boundary data are only available until November 2099; thus, in the period 2070–99, only 29 November and December values were applied. For application in the ANN, the second future period was shifted by one year. The ensemble approach results are used to construct conditional PDFs of changes in seasonal precipitation following the assumption that the climate response is an average of 30 differences, one per year, as discussed by Déqué (2009) and Déqué and Somot (2010). The model response R is calculated as
e1
where Xn is the seasonal mean of the year n and a is 60 for the period 2021–50 and 89 for the period 2070–99. The PDFs are based on bin counting, with a resolution of 0.12 mm day−1. The joint (mixed) PDF is
e2
where P denotes precipitation, PDFi is the PDF for precipitation of the model i, and pi is the probability of the model i. In the smoothing of the PDFs, the Gaussian kernel method has been applied.

e. Regional climate change projections

Multimodel combination is a pragmatic and accepted technique to improve the climate projections (Murphy et al. 2004; Rowell 2005; Tebaldi and Knutti 2007; Stott and Forest 2007; Déqué and Somot 2010). Table 3 summarizes the employed RCM models, GCM forcings, and the applied soil–vegetation–atmosphere transfer (SVAT) scheme. Four datasets available from GLOWA JR use the fifth-generation Pennsylvania State University–National Center for Atmospheric Research mesoscale model (MM5; Dudhia 1993), versions 3.5 and 3.7, driven with boundary data from the first realization of the ECHAM5 GCM and from the HadCM3 GCM. Evans (2009) examined the performance of 18 GCMs in the area of the Middle East. The results show that the GCMs applied in the present study perform reasonably well in both temperature and precipitation rate statistics. In reproducing the precipitation rate of the period 1979–2000, the ECHAM5 model revealed underestimation at the lower end of the model ensemble rather than being larger as in the HadCM3 global model, which was among the best performing models.

Table 3.

Regional data from the MM5 RCM runs and driving GCM used in the present study; r1 is the first realization of the ECHAM5 run, and Q0 is the first realization of the HadCM3 run. Runs ec5v35, ec5v37, hadv35, and hadv37 are ANN runs with RCM input. Here OSU stands for the Oregon State University land surface model.

Table 3.

An additional five datasets from the regional models Centre National de Recherches Météorologiques Regional Model, version 4 (CNRM-RM4), Regional Climate Model 3 (RegCM3), Regional Atmospheric Climate Model, version 2 (RACMO2), Community Land Model (CLM), and Hadley Center Regional Climate Model, version 3.0 (HadRM3.0), available from ENSEMBLES employ the third realizations of the ECHAM5 and HadCM3 and the Action de Recherche Petite Echelle Grande Echelle (ARPEGE) GCM (see Table 4). In the ENSEMBLES project, a larger number of RCM models were utilized. Only the models used in the present study cover the investigated area; the other models are close to the relaxation zone, which limits their performance and applicability. All employed models use the A1B Special Report on Emissions Scenarios (SRES; Nakicenovic 2000). The spatial resolution of the runs performed with two versions of the MM5 within the frame of GLOWA JR was 18.6 km and in the ENSEMBLES dataset was 25 km. A detailed description of the regional models employed in ENSEMBLES is given by Christensen et al. (2010) and of the MM5 used in GLOWA JR by Smiatek et al. (2011).

Table 4.

Regional data and driving GCM from the ENSEMBLES project used in the present study.

Table 4.

The present investigation employs bilinearly resampled versions of all available datasets, with a spatial resolution Δx of 0.25° compatible to the Climatic Research Unit (CRU) and E-OBS gridded observational reference data. In all statistics, an average area of 2Δx × 2Δx has been used. It covers large parts of the Figeh recharge area. An exact limitation of the investigation to the estimated recharge area is not possible with the available model resolution. As a prominent precipitation gradient is observed a very short distance from the Anti-Lebanon ranges to the Syrian Desert, resulting from a greater distance to the major moisture source, the Mediterranean Sea, and rain shadow effects, the model resolution may play a more important role. Investigations of Heckl (2011) and Gao et al. (2006) indicated that increased model resolutions can help to improve the model skills in the complex terrain of the EM region. Caldwell et al. (2009), however, conclude that simulated precipitation intensity is more related to processes internal to the model than to the grid spacing.

Smiatek et al. (2011) present a detailed evaluation of the MM5 runs, and an evaluation of the ENSEMBLES models for the present climate conditions has been presented by Klein Tank et al. (2009). The investigations show that a principal capability of the employed models is to reproduce the temperature and precipitation indices in the EM. However, the models still have problems in correct representation of the precipitation seasonality. They depict December and not January, as observed, as the month with the highest precipitation. To close gaps in observational reference data available for the investigated region, ENSEMBLES model runs in the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) were also used.

3. Results

a. Reproduction of present-day climate

For the period 1961–90, the ERA-40–driven simulations show an ensemble mean annual precipitation of 421 mm. The available observational reference, which, in principle, covers the southern two grid cells of the FI area in Fig. 1, depicts in the same period an annual mean precipitation of 364 mm. Limiting the ERA-40–driven results to this area yields 413 mm and an overestimation in the range of +15%. The simulations reproduce the precipitation seasonality and show good agreement in the winter and autumn seasons. In spring, especially in April and May, there is an overestimation present in the ERA-40 simulations.

The RCM-simulated ensemble mean annual precipitation in the area is 506 mm. Comparing the values of 417 mm from E-OBS, 421 mm from ERA-40–driven simulations, and 364 mm from the observations and taking into account a possible underestimation present in the observations, a moderate overestimation in the range of +20% can be assumed.

Observed and simulated seasonal values for the period 1961–90 are given in Table 5. For the winter season, the ERA-40–driven simulations of the five ENSEMBLES models yield in the investigation area an ensemble mean precipitation amount of 2.4 mm day−1, and the ensemble mean of the RCM models is 3.3 mm day−1. The interpolated observational precipitation from the E-OBS dataset is 3 mm day−1, while the mean value from the average station observation is only 2.1 mm day−1. The five MM5 RCM runs (see Table 5) show similar values, with slightly higher precipitation in the winter season. As no observations are available in the mountainous area, an underestimation in the observed reference can be assumed here. In total, the model ensemble reproduces the seasonal winter precipitation with some overestimation.

Table 5.

Observed and simulated mean seasonal temperature (°C) and precipitation (mm day−1) for the period 1961–90. ENS MM5 is the ensemble mean of the four MM5 models, and ENS all is the ensemble mean of all nine available models.

Table 5.

The values in intermediate seasons are similar, but monthly values, however, show differences in the precipitation seasonality. In the RCM ensemble, precipitation maximum occurs in December (see Fig. 6). Only two of the nine models depict the correct seasonality, but both show the largest deviations in the ensemble mean precipitation amount. Bias in precipitation seasonality can introduce substantial errors to any hydrological application. Therefore, a bias correction has been applied to the simulated precipitation data applied in the ANN. The correction procedure employs the quantile–quantile approach as shown by Déqué (2007). For each month and quantile, a correction factor is derived from the quotient of the 1961–90 mean modeled and observed data. This factor is then applied to all modeled daily precipitation values of the period 1961–2099 based on month and appropriate percentile.

Fig. 6.
Fig. 6.

Observed (OBS) and interpolated (E-OBS) data from the E-OBS dataset and simulated monthly ensemble mean (ENS) precipitation in the Figeh spring area.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

In ERA-40 simulations, the precipitation amounts in January (74.4) and February (74.5) are almost equal. Both observational references depict January as the month with maximum precipitation. The seasonal mean temperature agrees reasonably with the ERA-40–driven simulations but differs from the E-OBS gridded observations.

It is noted here that the temperatures were not corrected for differences in elevation between the observed and RCM-modeled grid cells. Taking into account the large temperature and precipitation gradients in the complex terrain of the Figeh spring system area, the available model ensembles reproduce the regional patterns with reasonable accuracy.

b. Future climate

Future climate simulations of the nine available regional climate models analyzed in this study indicate an increase in the annual mean temperature of +1.8°C for the period 2021–50 and +4.0°C for the period 2070–99 for the investigation area, both related to the 1961–90 annual mean. For the same periods, the average annual precipitation amount is simulated to decrease by −8.5% and −22%, respectively. All figures have been derived as ensemble mean values. Figure 7a shows the single-model temperature anomalies from the annual mean of the period 1961–90 for the Figeh area and the 10-yr running average of the ensemble mean value. The mean value has been calculated assuming equal weights for all nine models and shows a clear positive trend reaching a temperature increase of almost +4°C at the end of the century. Knutti et al. (2010) and Weigel et al. (2010) discuss the risk in model weighting in multimodel climate projections, recommending equal weighting as being safer and more transparent, as inappropriate weight may reduce the performance in cases where the internal model variability is large.

Fig. 7.
Fig. 7.

(a) Multimodel temperature and (b) precipitation anomalies from the 1961–90 mean. The thin red and blue lines show the single-model realizations; the black line shows the multimodel 10-yr running average.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

The single-model precipitation anomalies from the mean of the period 1961–90 are depicted in Fig. 7b. Here, the 10-yr running average of the ensemble mean shows an agreement with the observations of no particular trend in the period 1960–2000 and a small decreasing tendency since 2000. The negative trend amplifies after 2050.

The seasonal ensemble mean temperature increases for the period 2021–50, relative to the 1961–90 mean, are in the range of +1.4°C in winter to +2.2°C in summer. The figures for the intermediate season are +1.6°C in spring and +1.9°C in autumn. In the same period, the seasonal precipitation is simulated to decrease by −11% in winter, −8% in spring, and −2% in autumn. At the end of the century, the ensemble mean simulation indicates further temperature increases in the range of +3°C in winter to over +4.6°C in summer, as well as a precipitation decrease of −8% in autumn, −25% in winter, and −29% in spring.

Figure 8 illustrates the simulated ensemble mean changes in monthly temperature and snow cover. As the temperatures are simulated to increase in the range of more than +3°C in the winter months, a substantial decrease in the snow amount is expected. Consequently, the water storage potential will significantly decrease in both time and amount. Barnett et al. (2005) emphasize that, in snow-dominated regions, even without changes in precipitation intensity, temperature increases will lead to a shift in peak river runoff, and where storage capabilities are not sufficient, much of the runoff will be lost. In the Figeh area, the largest changes are simulated for the period 2021–50, where a decrease in the snow mass of more than −90%, as well as the temporal duration of the snow cover, is simulated to decrease by one month.

Fig. 8.
Fig. 8.

Simulated monthly ensemble (a) mean temperature and (b) snow cover in the periods 1961–90 (circles), 2021–50 (triangles), and 2070–99 (squares).

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Additional adverse effects on regional water availability are expected from warming-induced changes to evapotranspiration. On the other hand, pan evaporation observations in various countries show steadily decreasing values for the past 50 years (Barnett et al. 2005). Ohmura and Wild (2002) point out that a hemisphere evaporates more in winter than in summer, and there are large differences between the evaporation from land and the ocean, especially when land evaporation is from drying soil surface. Mariotti et al. (2008) investigated the Mediterranean water cycle changes by applying multimodel simulations. While the precipitation is projected to decrease throughout the year, the evaporation remains unchanged in the winter season and decreases in spring, summer, and autumn. In the result, the decrease in effective precipitation in the period 2070–99, relative to the 1950–2000 mean, is about −20%. In the eastern Mediterranean, a decrease in evaporation of the wet season is in the range of −0.05 to −0.1 mm day−1. From their investigations with ECHAM5 GCM, Al-Qinna et al. (2011) conclude a precipitation reduction of −10% for the Jordan rift valley. This is in the range of the present study. Figure 9 shows the ensemble mean changes in monthly cycles of evaporation E, precipitation P, and PE for the investigated periods 2021–50 (Fig. 9a) and 2070–99 (Fig. 9b). In the Figeh spring system area, a small evaporation increase is simulated for the late winter and early spring, reducing the effective precipitation. This trend is slightly amplified in the period 2070–99. In their investigations based on observation stations, LaMoreaux et al. (1989) found a moisture surplus for potential recharge, but only in the months from November to February. The RCM simulations indicate that this period might be longer. No effective changes are visible here in the period 2021–50; hence, changes to the precipitation amount and snow cover are the major drivers in the Figeh area.

Fig. 9.
Fig. 9.

Measurements of P, E, and PE. Solid lines represent 1961–90, dashed lines represent 2021–50, and dotted lines represent 2070–99.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Figure 10 shows the probability density functions for the simulated ensemble mean precipitation changes in the Figeh area in the autumn, winter, and spring seasons. The single PDFs of the model response R [see Eq. (1)] have been derived for all nine models. The mixed PDF has been calculated assuming equiprobable models, that is, the weighting factor pi = 1/9. Figure 10 reveals that major precipitation reduction is simulated for the winter and spring seasons, while the intermediate autumn remains almost unchanged. The data ensemble not may be fully sufficient for the purpose of delineation of full probabilistic climate change scenarios. This type of multimodel dataset is often described as an “ensemble of opportunity,” where various nonscientific aspects determine its composition, and thus, the sampling is neither systematic nor random (Tebaldi and Knutti 2007). Nevertheless, it is assumed to be the best presently available climate change data ensemble for the Figeh spring area. The probabilities of a precipitation increase of +1 mm day−1 and a precipitation decrease of −3 mm day−1 are equal in the winter season of the period 2021–50. At the end of the century, an additional shift toward negative precipitation changes is simulated. Even taking into account the limitations in the model ensemble composition, together with the decrease in snow mass, this is a reason for serious concern.

Fig. 10.
Fig. 10.

Probability density functions of the response value R for precipitation in the Figeh area for (a) 2021–50 and (b) 2070–99 relative to 1961–90.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

c. Hydrological climate change impact: Expected future Figeh spring discharge

The bias-corrected meteorological data of the four RCM model runs performed at KIT/IMK-IFU (see Table 3) were used as input into the developed ANN to assess the future climate change impact on Figeh discharge. The considered periods were the control period 1961–90 for the present-day climate and the periods 2021–50 and 2069–98 for the simulated future climate.

In comparison with measured Figeh discharge data of the control period, ANN runs with input from both HadCM3-driven RCMs depict lower-than-observed MQ peak flows. In contrast, the ECHAM5-driven simulation data reveal higher-than-observed peak flow values (not shown). The ANN runs of the period 1961–90 driven with the RCM data input yield an ensemble mean annual discharge of 194 × 106 m3 (see Table 6) with a range from 183 to 207 × 106 m3, documenting the good performance of the hydrological simulation system.

Table 6.

Observed and simulated annual discharge volumes in the investigated periods. Simulated discharge is ANN driven with observed meteorology.

Table 6.

Relative changes in the monthly MQ and MNQ values as simulated with input from the four RCMs are illustrated in Figs. 11a–d. With exception of the dry months of August and September, all simulations reveal partially substantial decreases. For the period 2021–50, the single realizations agree on the decreasing trend, with higher differences in the magnitude. At the end of the century, the differences between the single models are much smaller for both MQ and MNQ values. The ensemble mean of the relative change in MQ relative to the control period shows a decreased maximum during the peak flow from March to May, with values up to −20% in 2021–50 and almost −50% in 2069–98, relative to the 1961–90 mean. The course of the MNQ graph is similar to that of the MQ. MNQ statistics focus on low flow values, thus revealing higher reductions.

Fig. 11.
Fig. 11.

Change of (a) MQ 2021–50, (b) MQ 2069–98, (c) MNQ 2021–50, and (d) MNQ 2069–98 relative to 1961–90. Model acronyms are explained in Table 3.

Citation: Journal of Hydrometeorology 14, 2; 10.1175/JHM-D-12-065.1

Table 6 summarizes the ANN-simulated annual discharge volumes for the investigated periods 2021–50 and 2069–98 in comparison with the observed and simulated values for the period 1961–90. The results reveal that a moderate reduction of about −9% can be expected at the middle of the century. The reduction intensifies at the end of the century to −30%. Together with the expected population growth, this would significantly worsen the pressure on water availability in the region.

4. Conclusions

Simulation data from nine highly resolved regional climate models driven with boundary forcings from three different global models were used to assess the climate change signal in the Figeh spring area. The unique data ensemble is certainly still not sufficient for the purpose of delineation of probabilistic climate change scenarios; however, it is the best data ensemble available now, broadening significantly the scientific base in detection and attribution of possible climate change at regional scales. The investigation shows that the water supply from the spring system may face serious pressure under changed climate conditions in the already vulnerable arid and semiarid environment. Precipitation decreases about −8% in the period 2021–50 and −22% in the period 2070–99 and, together with increased temperatures of up to +3°C in winter, can substantially limit the water recharge potential already in the near future until 2050. Decreases in future snow coverage, today used as important water storage, and an increased percentage of liquid precipitation might lead to faster runoff and water losses on strong precipitation events. Additional adverse effects can result from increased interannual precipitation variability.

For the first time, a hydrological model able to describe the observed daily discharge by meteorological driving only was developed for the data-sparse and hydrogeologically complex Figeh spring. RCM data from four regional models were employed in the artificial neural network approach to simulate potential future discharge changes. All scenario projections depict a decrease in discharge intensity in the range of −9% to −30% for the investigated periods in the middle and end of the century. In the highly water-sensitive region of the Figeh spring area, this decrease in available water will likely have significant impacts on water security, human living, and economy. Adequate water management adaptation measures are therefore highly recommended.

Acknowledgments

The investigation was partially funded by the KfW Bankengruppe, Frankfurt am Main, Germany. ENSEMBLES data used in this work were funded by the European Union Sixth Framework Programme (EU FP6) Integrated Project ENSEMBLES (www.ensembles-eu.org; Contract 505539), whose support is gratefully acknowledged. We acknowledge the E-OBS dataset from the EU FP6 project ENSEMBLES and the data providers in the European Climate Assessment & Dataset (ECA&D) project (eca.knmi.nl) and the use of the R Package Kernel Smoothing (ks; see http://cran.r-project.org/web/packages/ks/index.html). The authors thank the Leibniz-Rechenzentrum, Garching, Germany, for providing the access to the HLRB II supercomputer.

APPENDIX

ANN Formulation

The ANN formulation is
ea1
where
  • Pa/b,5 = precipitation, 5-day window, lower/upper grid cell;
  • Pa/b,60 = precipitation, 60-day window, lower/upper grid cell;
  • em,60 = evapotranspiration index, 60-day window, grid cell average;
  • Pm,365 = precipitation, 365-day window, mean value;
  • b1(i) = bias for neurons i = 1:9 in the hidden layer;
  • w1(i, n) = the different weights for each dendrite n connected to the hidden layer neuron i = 1:9;
  • b2 = bias for the output layer neuron; and
  • w2(i) = the different weights for each dendrite i connected to the output layer neuron.
The parameters for weights and biases in transfer of the input to the hidden layer are shown in Table A1, and parameters for weights and biases in transfer of the hidden to the output layer are shown in Table A2.
Table A1.

Parameters for weights and biases in transfer input to hidden layer.

Table A1.
Table A2.

Parameters for weights and biases in the transfer from hidden to output layer.

Table A2.

REFERENCES

  • Al-Qinna, M., , Hammouri N. , , Obeidat M. , , and Ahmad F. , 2011: Drought analysis in Jordan under current and future climates. Climatic Change, 106, 421440, doi:10.1007/s10584-010-9954-y.

    • Search Google Scholar
    • Export Citation
  • Barnett, T. P., , Adam J. C. , , and Lettenmaier D. P. , 2005: Potential impacts of a warming climate on water availability in snow-dominated regions. Nature, 438, 303309, doi:10.1038/nature04141.

    • Search Google Scholar
    • Export Citation
  • Caldwell, P., , Chin H.-N. , , Bader D. , , and Bala G. , 2009: Evaluation of a WRF dynamical downscaling simulation over California. Climatic Change, 95, 499521, doi:10.1007/s10584-009-9583-5.

    • Search Google Scholar
    • Export Citation
  • Christensen, J., , Kjellström E. , , Giorgi F. , , Lenderink G. , , and Rummukainen M. , 2010: Weight assignment in regional climate models. Climate Res., 44, 179194.

    • Search Google Scholar
    • Export Citation
  • Déqué, M., 2007: Frequency of precipitation and temperature extremes over France in an anthropogenic scenario: Model results and statistical correction according to observed values. Global Planet. Change, 57, 1626, doi:10.1016/j.gloplacha.2006.11.030.

    • Search Google Scholar
    • Export Citation
  • Déqué, M., 2009: Temperature and precipitation probability density functions in ENSEMBLES regional scenarios. ENSEMBLES Tech. Rep. 5, 63 pp.

  • Déqué, M., , and Somot S. , 2010: Weighted frequency distributions express modelling uncertainties in the ENSEMBLES regional climate experiments. Climate Res., 44, 195209, doi:10.3354/cr00866.

    • Search Google Scholar
    • Export Citation
  • Dudhia, J., 1993: A nonhydrostatic version of the Penn State–NCAR mesoscale model: Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev., 121, 14931513.

    • Search Google Scholar
    • Export Citation
  • Evans, J. P., 2009: 21st century climate change in the Middle East. Climatic Change, 92, 417432, doi:10.1007/s10584-008-9438-5.

  • Evans, J. P., , Smith R. B. , , and Oglesby R. J. , 2004: Middle East climate simulations and dominant precipitation processes. Int. J. Climatol., 24, 16711694, doi:10.1002/joc.1084.

    • Search Google Scholar
    • Export Citation
  • Fleury, P., , Plagnes V. , , and Bakalowicz M. , 2007: Modelling of the functioning of karst aquifers with a reservoir model: Application to Fontaine de Vaucluse (south of France). J. Hydrol., 345, 3849, doi:10.1016/j.jhydrol.2007.07.014.

    • Search Google Scholar
    • Export Citation
  • Gao, X., , Pal J. S. , , and Giorgi F. , 2006: Projected changes in mean and extreme precipitation over the Mediterranean region from a high resolution double nested RCM simulation. Geophys. Res. Lett., 33, L03706, doi:10.1029/2005GL024954.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., , and Lionello P. , 2008: Climate change projections for the Mediterranean region. Global Planet. Change, 63, 90104, doi:10.1016/j.gloplacha.2007.09.005.

    • Search Google Scholar
    • Export Citation
  • Giorgi, F., , Bi X. , , and Pal J. S. , 2004: Mean, interannual variability and trends in a regional climate change experiment over Europe. I. Present-day climate (1961–1990). Climatic Change, 22, 733756, doi:10.1007/s00382-004-0409-x.

    • Search Google Scholar
    • Export Citation
  • Hansen, J., , Ruedy R. , , Sato M. , , and Lo K. , 2010: Global surface temperature change. Rev. Geophys., 48, RG4004, doi:10.1029/2010RG000345.

  • Haylock, M. R., , Hofstra N. , , Klein Tank A. M. G. , , Klok E. J. , , Jones P. D. , , and New M. , 2008: A European daily high-resolution gridded dataset of surface temperature and precipitation for 1950–2006. J. Geophys. Res., 113, D20119, doi:10.1029/2008JD010201.

    • Search Google Scholar
    • Export Citation
  • Heckl, A., 2011: Impact of climate change on the water availability in the near east and the upper Jordan River catchment. Ph.D. thesis, University of Augsburg, 181 pp. [Available online at http://opus.bibliothek.uni-augsburg.de/opus4/files/1603/dissertation_heckl.pdf.]

  • Hertig, E., , and Jacobeit J. , 2008: Downscaling future climate change: Temperature scenarios for the Mediterranean area. Global Planet. Change, 63, 127131.

    • Search Google Scholar
    • Export Citation
  • Hu, C., , Hao Y. , , Yeh T. , , Pang B. , , and Wu Z. , 2008: Simulation of spring flows from a karst aquifer with an artificial neural network. Hydrol. Processes, 22, 596604.

    • Search Google Scholar
    • Export Citation
  • Kafle, H., , and Bruins H. , 2009: Climatic trends in Israel 1970–2002: Warmer and increasing aridity inland. Climatic Change, 96, 6377, doi:10.1007/s10584-009-9578-2.

    • Search Google Scholar
    • Export Citation
  • Kattan, Z., 1997: Environmental isotope study of the major karst springs in Damascus limestone aquifer systems: Case of the Figeh and Barada springs. J. Hydrol., 193, 161182, doi:10.1016/S0022-1694(96)03137-X.

    • Search Google Scholar
    • Export Citation
  • Klein Tank, A., , Manzini E. , , Braconnot P. , , Doblas-Reyes F. , , Buishand T. , , and Morse A. , 2009: Evaluation of the ENSEMBLES prediction system. ENSEMBLES—Climate change and its impacts: Summary of research and results from the ENSEMBLES project, Final Rep., P. Van der Linden and J. Mitchell, Eds., Met Office Hadley Centre, Exeter, UK, 95–106. [Available online at http://ensembles-eu.metoffice.com/docs/Ensembles_final_report_Nov09.pdf.]

  • Knutti, R., , Furrer R. , , Tebaldi C. , , Cermak J. , , and Meehl G. A. , 2010: Challenges in combining projections from multiple climate models. J. Climate, 23, 27392758.

    • Search Google Scholar
    • Export Citation
  • Kunstmann, H., , Heckl A. , , and Rimmer A. , 2006: Physically based distributed hydrological modelling of the upper Jordan catchment and investigation of effective model equations. Adv. Geosci., 9, 123130.

    • Search Google Scholar
    • Export Citation
  • Kunstmann, H., , Suppan P. , , Heckl A. , , and Rimmer A. , 2007: Regional climate change in the Middle East and impact on hydrology in the upper Jordan catchment. IAHS Publ., 313, 141149.

    • Search Google Scholar
    • Export Citation
  • Kurtulus, B., , and Razack M. , 2010: Modeling daily discharge responses of a large karstic aquifer using soft computing methods: Artificial neural network and neuro-fuzzy. J. Hydrol., 381, 101111.

    • Search Google Scholar
    • Export Citation
  • LaMoreaux, P. E., , Hughes T. H. , , Memon B. A. , , and Lineback N. , 1989: Hydrogeologic assessment—Figeh spring, Damascus, Syria. Environ. Geol. Water Sci., 13, 73127, doi:10.1007/BF01664696.

    • Search Google Scholar
    • Export Citation
  • Lionello, P., , and Giorgi F. , 2007: Winter precipitation and cyclones in the Mediterranean region: Future climate scenarios in a regional simulation. Adv. Geosci., 12, 153158.

    • Search Google Scholar
    • Export Citation
  • Mariotti, A., , Zeng N. , , Yoon J. , , Artale V. , , Navarra A. , , Alpert P. , , and Li L. , 2008: Mediterranean water cycle changes: Transition to drier 21st century conditions in observations and CMIP3 simulation. Environ. Res. Lett., 3, 044001, doi:10.1088/1748-9326/3/4/044001.

    • Search Google Scholar
    • Export Citation
  • Meslmani, Y., 2009: Vulnerability assessment and adaptation measures to climate change in Syria. Final Rep., General Commission of Environmental Affairs, Damascus, Syria, 54 pp. [Available online at http://static.weadapt.org/placemarks/files/584/4dda6010f2e17INC-SY_V&A_General_Assessment_-En.pdf.]

  • Murphy, J. M., , Sexton D. M. H. , , Barnett D. N. , , Jones G. S. , , Webb M. J. , , Collins M. , , and Stainforth D. A. , 2004: Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature, 430, 768772, doi:10.1038/nature02771.

    • Search Google Scholar
    • Export Citation
  • Nakicenovic, N., , and Swart R. , Eds., 2000: Special Report on Emissions Scenarios. Cambridge University Press, 599 pp.

  • Ohmura, A., , and Wild M. , 2002: Is the hydrological cycle accelerating? Science, 298, 13451346, doi:10.1126/science.1078972.

  • Önol, B., , and Semazzi F. H. M. , 2009: Regionalization of climate change simulations over the eastern Mediterranean. J. Climate, 22, 19441961.

    • Search Google Scholar
    • Export Citation
  • Pöhler, H. A., 2006: Anpassung von WaSiM-ETH und die Erstellung und Berechnung von Landnutzungs- und Klimaszenarien für die Niederschlag-Abfluss-Modellierung am Biespiel des Osterzgebirges. Ph.D. thesis, Technische Universität Bergakademie Freiberg, 134 pp. [Available online at http://www.wasklim.de/download/Dissertation_Poehler.pdf.]

  • Rowell, D. P., 2005: A scenario of European climate change for the late twenty-first century: Seasonal means and interannual variability. Climate Dyn., 25, 837849, doi:10.1007/s00382-005-0068-6.

    • Search Google Scholar
    • Export Citation
  • Samuels, R., , Smiatek G. , , Krichak S. , , Kunstmann H. , , and Alpert P. , 2011: Extreme value indicators in highly resolved climate change simulations for the Jordan River area. J. Geophys. Res., 116, D24123, doi:10.1029/2011JD016322.

    • Search Google Scholar
    • Export Citation
  • Schulla, J., , and Jasper K. , 2007: Model description WaSiM-ETH. Tech. Rep., Institute for Geography, ETH Zürich, 181 pp.

  • Smadi, M., , and Zghoul A. , 2006: A sudden change in rainfall characteristics in Amman, Jordan during the mid 1950s. Amer. J. Environ. Sci., 2, 8491.

    • Search Google Scholar
    • Export Citation
  • Smiatek, G., , Kunstmann H. , , and Heckl A. , 2011: High-resolution climate change simulations for the Jordan River area. J. Geophys. Res., 116, D16111, doi:10.1029/2010JD015313.

    • Search Google Scholar
    • Export Citation
  • Soltani, S., , Saboohi R. , , and Yaghmaei L. , 2012: Rainfall and rainy days trend in Iran. Climatic Change, 110, 187213, doi:10.1007/s10584-011-0146-1.

    • Search Google Scholar
    • Export Citation
  • Somot, S., , Sevault F. , , Déqué M. , , and Crépon M. , 2008: 21st century climate change scenario for the Mediterranean using a coupled atmosphere–ocean regional climate model. Global Planet. Change, 63, 112126, doi:10.1016/j.gloplacha.2007.10.003.

    • Search Google Scholar
    • Export Citation
  • Sowers, J., , Vengosh A. , , and Weinthal E. , 2011: Climate change, water resources, and the politics of adaptation in the Middle East and North Africa. Climatic Change, 104, 599627, doi:10.1007/s10584-010-9835-4.

    • Search Google Scholar
    • Export Citation
  • Stott, P. A., , and Forest C. E. , 2007: Ensemble climate predictions using climate models and observational constraints. Philos. Trans. Roy. Soc. London, 365A, 20292052, doi:10.1098/rsta.2007.2075.

    • Search Google Scholar
    • Export Citation
  • Tebaldi, C., , and Knutti R. , 2007: The use of the multi-model ensemble in probabilistic climate projections. Philos. Trans. Roy. Soc. London, 365A, 20532075, doi:10.1098/rsta.2007.2076.

    • Search Google Scholar
    • Export Citation
  • Trichakis, I., , Nikolos I. , , and Karatzas G. , 2011: Artificial neural network (ANN) based modeling for karstic groundwater level simulation. Water Resour. Manage., 25, 11431152.

    • Search Google Scholar
    • Export Citation
  • Trigo, R. M., , Gouveia C. M. , , and Barriopedro D. , 2010: The intense 2007–2009 drought in the Fertile Crescent: Impacts and associated atmospheric circulation. Agric. For. Meteor., 150, 12451257, doi:10.1016/j.agrformet.2010.05.006.

    • Search Google Scholar
    • Export Citation
  • Weigel, A. P., , Knutti R. , , Liniger M. A. , , and Appenzeller C. , 2010: Risks of model weighting in multimodel climate projections. J. Climate, 23, 41754191.

    • Search Google Scholar
    • Export Citation
  • Xoplaki, E., , González-Rouco J. F. , , Luterbacher J. , , and Wanner H. , 2004: Wet season Mediterranean precipitation variability: Influence of large-scale dynamics and trends. Climate Dyn., 23, 6378, doi:10.1007/s00382-004-0422-0.

    • Search Google Scholar
    • Export Citation
Save