Journal of Hydrology 509 (2014) 193–206

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Multi-response calibration of a conceptual hydrological model in thesemiarid catchment of Wadi al Arab, Jordan

0022-1694/$ - see front matter � 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.jhydrol.2013.11.026

⇑ Corresponding author. Address: Helmholtz-Centre for Environmental Research,UFZ, Theodor-Lieser Str. 4, 06120 Halle, Saale, Germany. Tel.: +49 345 558 5208;fax: +49 345 558 5559.

E-mail address: tino.roediger@ufz.de (T. Rödiger).

T. Rödiger a,⇑, S. Geyer a, U. Mallast a, R. Merz a, P. Krause b, C. Fischer b, C. Siebert a

a Helmholtz-Centre for Environmental Research, UFZ, Germanyb Friedrich-Schiller University Jena, Geography, Germany

a r t i c l e i n f o s u m m a r y

Article history:Received 29 April 2013Received in revised form 23 October 2013Accepted 16 November 2013Available online 1 December 2013This manuscript was handled byKonstantine P. Georgakakos, Editor-in-Chief,with the assistance of Marco Borga,Associate Editor

Keywords:Groundwater rechargeHydrological modellingMulti-response calibrationJ2000gSemiaridJordan

A key factor for sustainable management of groundwater systems is the accurate estimation of ground-water recharge. Hydrological models are common tools for such estimations and widely used. As suchmodels need to be calibrated against measured values, the absence of adequate data can be problematic.We present a nested multi-response calibration approach for a semi-distributed hydrological model inthe semi-arid catchment of Wadi al Arab in Jordan, with sparsely available runoff data. The basic ideaof the calibration approach is to use diverse observations in a nested strategy, in which sub-parts ofthe model are calibrated to various observation data types in a consecutive manner. First, the availabledifferent data sources have to be screened for information content of processes, e.g. if data sources con-tain information on mean values, spatial or temporal variability etc. for the entire catchment or only sub-catchments. In a second step, the information content has to be mapped to relevant model components,which represent these processes. Then the data source is used to calibrate the respective subset of modelparameters, while the remaining model parameters remain unchanged. This mapping is repeated forother available data sources. In that study the gauged spring discharge (GSD) method, flash flood obser-vations and data from the chloride mass balance (CMB) are used to derive plausible parameter ranges forthe conceptual hydrological model J2000g. The water table fluctuation (WTF) method is used to validatethe model. Results from modelling using a priori parameter values from literature as a benchmark arecompared. The estimated recharge rates of the calibrated model deviate less than ±10% from the esti-mates derived from WTF method. Larger differences are visible in the years with high uncertainties inrainfall input data. The performance of the calibrated model during validation produces better resultsthan applying the model with only a priori parameter values. The model with a priori parameter valuesfrom literature tends to overestimate recharge rates with up to 30%, particular in the wet winter of 1991/1992. An overestimation of groundwater recharge and hence available water resources clearly endangersreliable water resource managing in water scarce region. The proposed nested multi-response approachmay help to better predict water resources despite data scarcity.

� 2013 Elsevier B.V. All rights reserved.

1. Introduction

Groundwater resources in Jordan are strongly limited due to(semi-) arid climate conditions. Significant overuse by groundwa-ter abstraction intensifies the situation, clearly documented by fall-ing groundwater levels, the disappearance of springs and saltwaterintrusions from deeper aquifers.

For sustainable future use of water resources reliable estimatesof water balance components are necessary, particularly ofgroundwater recharge as the main source of replenishment ofwater resources. Linking between hydrological and groundwater

flow models is an appealing way to estimate groundwater rechargerates in semi-arid regions. The spatio-temporal patterns of GW re-charge as output of the hydrological models is thereby used as in-put in GW flow models for managing water resources (e.g. Al-Abedet al., 2005; Abdulla and Al-Assa’D, 2006; Wu et al., 2011; Gräbeet al., 2012). Hydrological models may be seen as a transformatorof input, e.g. rainfall to output, e.g. runoff. This transformation fol-lows (a) some general (physical or conceptual) principles, i.e. themodel structure or model equations, and (b) some degree of free-dom to adapt the general principles to the local conditions, themodel parameters, which are calibrated. The hydrological modelhas to be complex enough to account for the dominant process,but simple enough not to be over-parameterized (Blöschl andGrayson, 2002).

However, complex highly non-linear processes and sparse datachallenge hydrological modelling in semi-arid regions as Jordan.

194 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

Model of intermediate complexity, such as J2000 g (Kralisch andKrause, 2006) has to be proven to be a good balance to accountfor the complex and the scarcity on data to reliably estimate modelparameters. Similar to the choice of model structure, data scarcityalso limits classical approaches to calibrate model parameters (e.g.Klemeš, 1986). One approach to overcome the limitation of sparsecalibration data is to use additional data sources to calibrate or val-idate model parameter sets. These can be geo-chemical data(Mroczkowski et al., 1997), groundwater data (Madsen, 2003), soilmoisture data (Western and Grayson, 2000) or regional relation-ship of model parameters (Parajka et al., 2007a,b). The general ideaof these approaches is to constrict the model parameter space byestablishing a multi-objective calibration, i.e. by trying to fit modelstate variables, such as soil moisture or groundwater recharge toobservations, additional to the classical calibration towards runoff.For example Marce et al. (2008) successfully use a multi-objectivefunction calibration strategy to solve the parameterization of acomplex application of a hydrological model.

Such additional data may be available in various degrees of spa-tio-temporal information content, such as time series of hydrolog-ical variables at single locations within the catchment, e.g. timesseries of groundwater levels in a well, or as integral measures overtime and space, such as groundwater recharge derived from chlo-ride mass balance. This clearly affects the approach to calibratemodel parameters. A common way is to calibrate all model param-eter at once, using a multi-objective optimization function, to fitmodel parameters to the times series of observed runoff and e.g.soil moisture data. However, this requires spatio-temporal infor-mation content of the additional data source. If only integral infor-mation in space and time are available, different strategies areneeded.

First, more than one additional source of data may be used. Forexample, Sharda et al. (2006) uses chloride mass balance (CMB)and water table fluctuation (WTF) in a multi-response approachfor the quantification of the groundwater recharge. Sophocleous(1991) successfully combines a soil water model and the water ta-ble fluctuation method to quantify groundwater recharge. A sec-ond aspect concerns the calibration of model parameters toadditional data. One promising strategy is to use a nested ap-proach, in which some sub-parts of the model are calibrated tothe data in a consecutive manner. Such sub-parts can be sequencesof hydrological processes, e.g. parameters of the soil moisture rou-tines are calibrated first to measured soil moisture data, then allremaining parameters are calibrated to runoff, while soil moistureparameters remain unchanged in the second step. The idea of thisapproach is to reduce the degree of freedom. If all parameters arechanged at once, unrealistic values of one parameter can be com-pensated by other parameters and, hence, a large amount of differ-ent parameter sets give equally good performance. This effect ofequifinality is widely shown in hydrological modelling (e.g. Bevenand Binley, 1992). Independent calibration of some parameters in aconsecutive manner may reduce compensation.

The objective of this study is to apply a nested multi-responsecalibration approach for the conceptual hydrological modelJ2000 g to estimated groundwater recharge in the semi-arid WadiAl-Arab catchment (Jordan). Information on groundwater rechargebased on the chloride mass balance (CMB), observed spring dis-charges and surface runoff observations are used for calibration.The calibrated model is then validated using the water table fluc-tuation method and compared to estimates given by the model,with a priori parameter values derived from literature values.

2. Study area of Wadi al Arab

Wadi al Arab is a semi-arid catchment in NW Jordan (Fig. 1) lo-cated between the Yarmouk River valley in the north, the Jordan

River Valley in the west, the foothills of the Ajloun Dome in thesouth and the branch of the Azraq plain in the east. The altitudevaries from �20 m mean sea level (msl) in the Jordan River Valleyto more than 1100 m msl in the mountain range of the Ajloun. Thesurface catchment area covers about 200 km2 (Fig. 1B), while thegroundwater catchment contains an area of about 300 km2

(Fig. 1A). Important perennial springs exist only in the north-wes-tern part of the study area (Fig. 1A and C). Wadi al Arab starts toincise south of Irbid, the second largest Jordan city and continuesNW-wards. South of the perennial spring region, the Wadi abruptlychanges its flow-direction to the SW (Fig. 1).

The annual precipitation varies between <350 mm and>550 mm (Ref. to Fig. 3A). Precipitation occurs generally in thewet season from October to April, while the largest quantities withup to 80% fall between December and February. Average daily tem-perature varies from about 12.5 �C during winter season (fromNovember to April) to around 23 �C during summer season (fromMay to October).

Alternating sequences of Cretaceous to Cenozoic age and differ-ent hydraulic conductivities represent the geological setting of thecatchment (Fig. 1). The fractured and karstified package from theCretaceous Upper Ajloun Group (A7) to the Cenozoic Lower BelqaGroup (B1 + B2) is the major aquifer for water supply in the studyarea. On its top, the A7/B2 aquifer is hydraulically separated fromthe locally productive B4 aquifer by the Muwaqqar aquitard (B3).Due to the NW-wards dip of the formations, the A7/B2-aquifer be-comes confined in the NW part of the study area.

As a follow up of the geological and climatic conditions, themain soil types are Inceptisols (Terra Rossa and Dark Rendzina),Entisols and Vertisols (Grumusol) (USDA, 1975). The depth of thecorresponding soil types shows a strong variation and dependsstrongly on the dominating vegetation cover. Predominant soiltypes feature A- and C-horizons with a mis-sing B-horizon. In areaswith poor vegetation and steep slopes even the A-horizon is mis-sing. During rain fall events the dry and immature soils possessonly small capabilities to store an adequate amount of water. Thus,in combination with the steep morphology the precipitation excessaccumulates quickly and flows downhill inducing flash floods.

According to Zohary (1973), Wadi al Arab belongs to the Medi-terranean and the Irano-Turanian vegetation zones. The Mediterra-nean vegetation zones are characterized especially by Pinushalepensis, Quercus coccifera, Quercus ithaburensis, Ceratonia siliqua,Olea europaea and Pistacia spp. The Irano-Turanian zone consists ofmostly shrubs and bushes, like Retama raetam, Ziziphus lotus, Arte-misia herba-alba, Noaea mucronata and Anabasis syriaca (Al-Eisawi,1996).

3. Methods

3.1. GSD – gauged spring discharge method

The springs of Ein Umm Qais (catchment 1.7 km2) and Ein elAsal (catchment 1.1 km2) drains the two corresponding headwatercatchments. Spring discharges were manually measured fromOctober 2007 to October 2008. Fig. 2 shows the discharge curvesof both springs versus precipitation between October 2007 andOctober 2008. Both springs quickly react on the precipitation ofthe rainy season in Winter 2007/2008. However, Ein el Asal re-sponses slightly earlier than Ein Umm Qais and shows a muchlonger lasting discharge peak up to the end of the wet season.The direct response of the spring discharges on the precipitationevents and the continuing spring discharge in the dry summer sea-son indicates that the discharging aquifer is characterised by adouble porosity system that is dominated by matrix flow. The aver-age spring discharges are in Ein Umm Qais 0.41 l/s and in Ein el

Fig. 1. (A) The extent of the groundwater catchment area (black line), (B) extent of the surface catchment area, (C) the position of the perennial springs (Ein Umm Qais (gwcatchment area black line), Ein el Asal (gw catchment area black line) and Ein Khanzir), the observation well Kufr Asad 3 and the built-up area of Umm Qais are shown.

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 195

Asal 0.53 l/s. Assuming the individual integrated spring dischargeto be equal to the groundwater recharge, the resulting estimatedgroundwater recharge is 26 mm/yr in Ein Umm Qais and 58 mm/yr in Ein el Asal catchment, respectively. The large differences re-sult from large occurrence of impervious basalts as top layer inUmm Qais catchment.

3.2. CMB – chloride mass balance method

The chloride mass balance method is a simple method to esti-mate the groundwater recharge in arid and semi-arid areas (Bazu-hair and Wood, 1996; Sharda et al., 2006). It is based on thetemporal and spatial distribution of groundwater recharge usingthe conservative chloride (Cl) concentration in groundwater. Ahandful of assumptions are necessary for a successful utilization:e.g. (1) chloride is conservative in the system, (2) steady state con-ditions are maintained in regarding to long term precipitation andchloride concentration in that precipitation, (3) total chloridedeposition (precipitation and dry fallout) at land surface is the onlysource of chloride in groundwater represents, (4) no surface runoffdischarges occurs beyond the boundaries of the aquifer, (5) noevaporation of groundwater occurs upgradient from groundwater

sampling points (Bazuhair and Wood, 1996). According to the totalchloride depositions at land surface an important source of uncer-tainty for applying the CMB method is that only limited historicaldata in the rates are available and networks for chloride depositionin dry fallout are sparsely or not existent. These fact leads to anunderestimating of the total chloride deposition at land surfaceand consequently to an underrating of groundwater recharge byusing the CMB method. A comprehensive geological/geochemicalmapping of the catchment did not reveal additional natural Cl-sources, although eolian deposition of Cl-containing sedimentscannot be excluded.

Hence, recharge R is calculated according to Eq. (1)(Wood andSanford, 1995) by taking the average annual amount of precipita-tion P (mm) and the average Cl-concentration in precipitation(Clrain) and in groundwater (ClGW) of the investigated aquifer intoaccount:

R ¼ P � Clrain=ClGW ð1Þ

During field campaigns between 2007 and 2009, 28 groundwa-ter samples and one cistern (precipitation) sample was taken andanalysed by ionchromatography (IC). Additional values for Cl inprecipitation were gathered from Kattan (2006) and Bajjali

Fig. 3. (A) Interpolated Cl-concentrations (black numbers [mg Cl/l]) and average precipitation amount (white numbers [mm/yr]) and (B) interpolation of Cl-concentrations[mg/l] in springs and wells in the study area.

Fig. 2. Relation between spring discharge of Ein Umm Qais (black) and Ein el Asal (grey) and rain events, measured at the climate station in Umm Qais village.

196 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

(2006). The precipitation average was calculated by interpolatingbetween available rain stations (Table 1) in the vicinity of the studyarea. All data were used along with inverse distance weightedinterpolation technique to generate an iso-concentration map ofCl-concentration in precipitation and in the main aquifer A7/B2(Fig. 3). Fig. 3A and Table 2 shows typical mean Cl contents (ca.6 mg/l) and average amounts (470 mm/a) of precipitation for thestudy area. Fig. 3B shows the Cl concentrations for the groundwa-ter samples in the study area, which range from 25.5 mg/l to158 mg/l. Cl concentrations of >100 mg/l in groundwaters of A7aquifer at the western edge of the study area (North Shuneh) areaccompanied by significantly higher temperatures (55 �C) and spe-cific hydrochemical compositions indicating the ascension of deepseated groundwater into the A7/B2 aquifer. For that reason thesewater samples were excluded for the estimation of the mean Cl-concentration (41 mg/l) in A7/B2. Using this concentration agroundwater recharge for A7/B2 was calculated to be about63 mm/yr.

Average Cl-concentration of springs Ein Umm Qais and EinKhanzir water (both B4 aquifer are 88 mg/l and 158 mg/l, respec-tively (Fig. 3B), presenting the result of an intense contact to

basalts in their catchments. Annual groundwater recharge ofaround 20–30 mm is estimated by the CMB for both springs. Incontrast annual groundwater recharge of around 60 mm are calcu-lated for the spring Ein el Asal (limestone aquifer), which is charac-terized by average Cl-concentration of 32.4 mg/l. The analysedranges of groundwater recharge are comparable to the above sta-ted GSD-results for Ein Umm Qais and Ein el Asal.

3.3. WTF – water table fluctuation method

The water table fluctuation (WTF) method is a simple methodto derive groundwater recharge estimates from observed risinggroundwater tables. A detailed description of the approach is givenin Healy and Cook (2002). The basic idea of the method is thatgroundwater recharge and net subsurface flow, including baseflow,evaporation from groundwater and groundwater extraction, deter-mine the change in stored water in the aquifer and hence thechange in the water table. The WTF method is based on the pre-mise that a time lag occurs between the arrival of recharge waterin the aquifer inducing a rises of the groundwater table and theredistribution of that water to net subsurface flow, such as

Table 1Climate stations and respective parameters that were used for the model setup (fortime-series the maximum range is given; temporal gaps are not shown).

Location Parameter Time period Unit

Irbid (AE 0001)b Precipitation 1937–2005 mm/dIrbid (AE 0002)b Precipitation 1954–2007 mm/dKufr Yuba (AE 0003)b Precipitation 1937–2006 mm/dKufr Asad (AE 0004)b Precipitation 1962–2006 mm/dAl-Taiyiba (AB 0001)b Precipitation 1937–2006 mm/dDeir Abi Said (AB 0002)b precipitation 1937–2006 mm/dUmm Quais (AD 0005)b Precipitation 1937–2006 mm/dKharja (AD 0008)b Precipitation 1937–1999 mm/dHusn (AD 0010)b Precipitation 1995–2006 mm/dIbbin (AD 0018)b Precipitation 1995–2006 mm/dBaqura (AD 0032)b Precipitation 1995–2006 mm/dRihaba (AF 0002)b Precipitation 1995–2004 mm/dRas Muneef (AH 003)b precipitation 1995–2005 mm/dUmm Quaisa Climate 2007–2009 mm/dIrbidc Climate 1980–2009 mm/dQueen Allia airportc Climate 1980–2009 mm/dBet Sheand Climate 1961–2000 mm/d

a Own data.b Ministry of water and irrigation (available through DAISY: www.ufz.de/

index.php?de=16882).c Wunderground web-service: www.wunderground.com; Tutiempo:

www.tutiempo.net.d Metbroker: http://pc105.narc.affrc.go.jp/metbroker/.

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 197

baseflow. Within that time lag the change in water table in uncon-fined aquifers can be accounted to recharge only and recharge canbe calculate according to

R ¼ S � DhDtA ð2Þ

where R is groundwater recharge, S is storage coefficient (effectiveporosity in unconfined area) or specific yield, Dh is the change ingroundwater table in the period Dt and A is catchment area. Thelength of the time lag and an uncertain storage coefficient is criticalin the success of this method. For the local aquifer system an aver-age storage coefficient S = 0.017 was estimated based on pumpingtests and reports in BGR and WAJ (1995, 1996, 2001).

According to Healy and Cook (2002) the method has its limita-tions but can be best applied to water tables that display sharpwater-level rises and declines. This is observable at Kufr Asad 3observation well, situated in the western wadi part. In that well,groundwater levels were continuously measured monthly 1–2times from 1983 to 2003 (Fig. 4). The temporal resolution is as-sumed to be sufficient due to a strong seasonal pattern of winterrainfall with subsequent periods of rising water tables and drysummer periods with declining water tables. The closer surround-ing of observation well Kufr Asad 3 is free of production wells, a lo-cal effect of water extraction is not visible.

Groundwater levels in the local aquifer are generally decliningfrom 1982 to 1991 (Fig. 4). Although the abstraction in the Kufr

Table 2Chloride concentrations in precipitation for selected stations used for calculating groundw

Station Coordinates (UTM, zone 36 N; WGS 1984) C

East North M

Ras Muneefa 758746 3584775 4Irbida 767685 3602767 5Deir Allaa 746002 3566701 2Kounietrab 739976 3602767 2Izraab 795907 3602451 2Suwedab 823269 3599967 2Cisternc 760059 3614857 4.

a Bajjali (2006).b Kattan (2006) -GNIP-stations: IAEA/WMO (2006).c Own measurement.

Asad- and Wadi Al Arab well fields (situated in the down gradientarea of the observation well Kufr Asad 3) were reduced at the endof the 1980s, groundwater level in Kufr Asad 3 showed only a shortand light recovery. It is also observable that the groundwater levelin Kufr Asad 3 well reacts with a short delay to seasonal rechargeevents. Moreover, since the observation well is located close tothe outlet of the catchment, it reflects all hydraulic reactions andchanges of the entire system. By applying WTF in Kufr Asad 3 well,the estimated groundwater recharge varies in a range of 12 mm(1984) to 230 mm (1992) with an average of 65 mm/yr, calculatedfor the period 1983–1998.

The groundwater recharge derived from the WTF method issimilar to the estimates of the CMB method (63 mm/yr) and theGSD method (58 mm/yr) for Ein el Asal. The fact that three inde-pendent methods result in a similar range of groundwater rechargeindicates the plausibility of the groundwater recharge estimate.Due to the abundant basalt covers west of Umm Qais, which obvi-ously hinder infiltration and reduce groundwater recharge, GSD-based recharge rates of Ein Umm Qais are much smaller, about26 mm/yr.

3.4. Hydrological model J2000g

The hydrological model J2000 g is a rainfall-runoff model toestimate spatially distributed hydrological water balance compo-nents (Krause and Hanisch, 2009; Krause et al. 2010). J2000 g isimplemented within the modular oriented JAVA framework sys-tem JAMS (Jena Adaptable Modelling System) (Kralisch and Krause,2006). Fig. 5 illustrates the model input requirements in additionto the simulated processes and the related output data.

The model requires daily or monthly meteorological input data(rainfall, temperature, sunshine duration, relative humidity, windspeed) and spatially distributed information on topography, soiltypes and land-cover to describe the physio-geograpcical condi-tions of the study area. The catchment is distributed in hydrologi-cal homogenous entities, so called Hydrological Response Units(HRU). HRUs are delineated by GIS overlay analysis of relevant spa-tially distributed information. Based on GIS derived informationeach HRU is described by its elevation, slope, aspect, land-coverand soil type

For each HRU and each time step, water balance components i.e.evaporation are calculated according to the climatological inputdata, which are regionalized by spatial interpolation using in-verse-distance-weighting and optional elevation correction. Poten-tial evapotranspiration (PET) is calculated using the Penman–Monteith formula as described in Allen et al. (1998). To accountfor uncertain input data a multiplicative calibration parameter(b) can be used to increase or decrease the potential evapotranspi-ration for all HRU by the same relative amount.

The central element of the J2000 g water balance is the soilwater storage capacity (SWS). The SWS handles the distribution

ater recharge according to the CMB method.

l-conc. of rainfall (mg/l) Average precipitation (mm/yr)

in Max Mean

17 10.6 58212 6.8 47317 6 28510 6 16810 4 11512 4.8 159

6 6 5.3 –

Fig. 4. Groundwater levels [m] of Kufr Asad 3 well, monthly precipitation [mm/month] and the calculated groundwater recharge using the WTF-method [mm/yr].

Fig. 5. Schematic diagram illustrating the J2000g model input requirements, the simulated processes and the related output data.

198 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

between input (i.e. precipitation) and output (evaporation, directrunoff, groundwater recharge and baseflow.) The maximum soilwater storage capacity (maxSWS) is parameterised using the effec-tive field capacity (FC) of the soil horizons within the rooting zone(depth) RD. With the calibration parameter FCA the maximum soil

water storage capacity (maxSWS) can be increased or decreased bythe same relative amount for all HRUs.

The input i.e. precipitation (P) is used until the complete satura-tion of maxSWS. After saturation, the excess water (EW1) will beallocated to the evaporation. The remaining portion of EW1 that

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 199

gets not evaporated will be taken as input (EW2) for the generationof direct runoff and groundwater recharge. In the case that EW1 issmaller as the potential evaporation (PET) an evaporation deficitexists that is partly balanced by SWS. Therefore a maximum evap-oration rate (ER) from the SWS is calculated on the ratio of a linearcalibration coefficient (ETR) and the relative saturation of soilwater storage (Eq. (3)).

ER ¼ DSWS=ETR �maxSWS ð3Þ

The excess water (EW2) is portioned into a direct runoff compo-nent (DQ) and groundwater recharge (GR) based on the slope (a)and a calibration coefficient LVD (between 0 and 1) (Eqs. (4) and(5)).

DQ ¼ tan a � LVD � EW2 ð4Þ

GR ¼ ð1� tan aÞ � ð1-LVDÞ � EW2 ð5Þ

In the framework of the groundwater module the generated GRgets partitioned into two groundwater reservoirs (GWS1, GWS2),realised by the calibration coefficient c.

The two-groundwater reservoirs allow the simulation of a fastand slow groundwater component (double continuum aquifer).The outflow (baseflow) from GWS1 and GWS2 is computed withsingle linear storage cascades (Nash, 1958) that are parameterisedby n linear reservoirs and the retention coefficient k. The retentioncoefficient k was estimated by Eq. (6):

k ¼ KD=n ð6Þ

The delay time KD is estimated from the falling limb of a mea-sured discharge hydrograph via Eq. (7). Qi and Qi+1 describe a seriesof discharge measurements on the decreasing part of the hydro-graph in the time step Dt.

KD ¼ Dt=ðlnðQ i=Q iþ1ÞÞ ð7Þ

3.5. Data and model setup

To run the hydrological model, monthly data series of precipita-tion, relative humidity, relative sunshine duration, and tempera-ture and wind velocity were available for the period January1980 to December 2008. All input data are given in Table 1.

The total observed discharge of 6 flash flood events were onlyavailable for the period 2000–2005 (data provided by the Ministryof Water and Irrigation, Jordan).

For the spatial discretisation of the catchment a mesh of irreg-ular triangular HRU elements with a mesh base length between 50and 400 m was generated. Based on that mesh, spatially distrib-uted parameters describing the HRU were derived from GIS alongwith a priori values from literature (see below).

Topographical parameters (slope, aspect and elevation) wereobtained from a digital elevation model (DEM 25 � 25 m). Accord-ing to Tilch et al. (2002), the slopes (s) and their according surfaceratios (A) were classified into four groups: (i) 0� < s 6 5�, A = 0.14;(ii) 5� < s 6 15�, A = 0.51; (iii) 15� < s 6 30�, A = 0.27 and (iv)30� < s 6 90�, A = 0.8. The area aspects and their according A wereclassified into eight classes by 45� segments and one class by �1:NNE, A = 0.17; ENE, A = 0.87; ESE, A = 0.44; SSE, A = 0.69; SSW,A = 0.14; WSW, A = 0.15; WNW, A = 0.15; NNW, A = 0.15; and �1,A = 0.58).

Land-cover was classified by using an ASTER image from May2000. Ground-truth data were obtained from different field cam-paigns during 2007 and 2009. According to the dominant planttypes (Table 3), 8 major and 13 sub classes were distinguished.Vegetation specific parameters like leaf area index or stomataresistance were estimated from literature values (Wilson and Hen-derson-Sellers, 1985; Dorman and Sellers, 1989; Rhizopoulou and

Mitrakos, 1990; Acherar and Rambal, 1992; Körner, 1994; Schulzeet al., 1994; Kergoat, 1998; Baldocchi et al., 2004; Niinemets et al.,2005) shown in Table 3.

From the National Soil Map of Jordan (Ministry of Agriculture,Jordan, 1994) 8 main soil classes and 20 subclasses were digitized.These classes can be distinguished by grain size, porosities, fieldcapacities and thickness (Table 4). The maximum percolationcapacities for the geological units were derived from Berndtssonand Larson (1987). As stated above, to generate the HRU, all inputparameters were spatially integrated, which resulted in a spatiallydiscriminated mesh of 23,367 HRU-cells.

3.6. Calibration/validation strategy

A classical runoff calibration and validation approach (see e.g.Klemeš, 1986) is not possible due to the absence of observed timeseries of runoff. Given the fact that different data types (e.g.groundwater recharge rates based on the CMB data or single runoffobservations) with a limited information value (e.g. mean valuesvs. detailed times series) are only available for calibration and val-idation, a nested multi-response calibration strategy is applied. Theidea of the approach is to use diverse observations in a nestedstrategy, in which sub-parts of the model are calibrated to variousobservation data types in a consecutive manner (Fig. 6). An impor-tant step is the screening of the information content of availabledata types to identify which data types can be used to derive real-istic ranges for single model parameters or a subset of modelparameters. The mapping of which data source is used to calibratewhich model parameter is of course site dependent, depending onavailable data and on the used model. Hence the following nestedmulti-response calibration approach is site dependent, but mayserve as a general example.

For the given catchment the approach is based on the followinghydrological assumptions and the resulting calibration steps(Fig. 6):

(1) In a first step, parameter values from literature are used tobuild up an a priori model with best guess parameter esti-mates. FCA is assumed to be equal to field capacity form soildata. The default settings of FCA, ETR, LVD, b were set by 1 touse the initial values and run the model without calibration.

The coefficients n and k are set according to Amraoui et al.(2013) with KD1 = 5 for a well-drained karst aquifer andKD2 > 50 days for aquifers having a considerable storage capacity.The splitting coefficient c between the two-groundwater storagesis chosen to be in the range of 50% to 75%.

(1) Spring discharges in the Ein Umm Quais and Ein el Asal rep-resent the baseflow dynamics of the corresponding headwa-ters. Hence observed spring discharges can be used tocalibrate those model parameters, which reflect the base-flow component in the headwaters. In J2000 g, baseflow iscontrolled by the parameters n (number of reservoirs in aNash Cascade), k (retention coefficient) and c which dividesthe recharged water to the two groundwater reservoirs. Fol-lowing this assumption J2000 g is sued to simulate baseflowof headwaters by calibrating n, k and c only.

The so calibrated parameters of the headwaters are then as-sumed to be valid for the entire catchment. For the given catch-ment the baseflow dynamics of spring of Ein el Asal is assumedto be representative for the baseflow dynamics of the entire catch-ment, as it sufficiently represents the limestone character of Wadial Arab. A recession curve analysis indicates that two baseflowcomponents are dominating, a fast baseflow component with KD

Table 3Land cover parameters, used for the hydrological simulation.

Land cover/subclasses Area % Surface resistance of the land cover (s/m) (literature) Leaf Area Index (m2/m2) (literature)

Impervious areas (dense + light) 2.8 10 –Agriculture (cultivated + noncultivated) 23.8 50–1000 0–1Water 0.2 10 –Rocks 4.7 10 –Shrubs 3.3 102–323 4–5Forest (coniferous + deciduous) 1.6 118–556 4–7Rangeland (grass/hill grass (very light)/bare soil) 60.0 60–1000 0–1.6Olive orchards 3.6 150–1250 3

Table 4Soil parameter used for the hydrological simulation and derived from the NationalSoil Map of Jordan (Ministry of Agriculture, Jordan, 1994).

Soil class/amount subclasses Area % Depth(cm)

Field capacity(mm)

Vertisols; Entic Chromoxererts/2 4.87 98–110 178.5–200.7Inceptisols; Typic Xerochrepts/10 63.24 50–120 97.2–212.2Inceptisols, Lithic Xerochrepts/1 7.81 60 101.4Inceptisols, Vertic Xerochrepts/1 0.42 100 182.5Inceptisols, Calcixerollic Xerochrepts/1 3.47 80 189.9Entisols, Lithic Xerorthents/3 17.95 40–70 73.0–162.0Ardisols, Ustochreptic Camborthids/1 0.52 90 164.2Ardisols, Ustollic Camborthids/1 1.72 90 175.5

200 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

to be mainly between 5 and 10 days and a much slower componentwith KD 6 90 days.

To start with a model as simple as possible, it is assumed in afirst step, that only one groundwater storage, simulated in theJ2000 g model with one Nash cascade, dominates the baseflow ofthe headwaters. In Fig. 7 simulated baseflow vs. observed springdischarges are given for different parameter combination for onefast baseflow component. Each scatter-plot diagram represents

Fig. 6. Workflow of the multi-response calibratio

different combinations of k and n and the associated R2 for simu-lated and measured data, while each symbol in the scatter plot rep-resents the spring discharges of 1 month. Using only one fastbaseflow component the best two correlations are obtained for k1.75/n 4 (KD = 7, R2 = 0.818) and k 1.6/n 5 (KD = 8, R2 = 0.815), how-ever the simulated discharges are always smaller than observeddischarges.

To overcome this problem a second groundwater storage isintroduced, representing much slower baseflow components. Suchdual groundwater component dynamics are typical for doubleporosity (karstified) aquifer systems, where groundwater is re-charged by (1) a fast component, using fractures and fissures asflow path (k1/n1) and (2) a slow matrix flow component (k2/n)(Rödiger et al., 2009). Fig. 8 shows simulated baseflow vs. observedspring discharges for different parameter combination of a modelwith one fast and one slow reacting groundwater storage. Eachscatter-plot diagram represents different combinations of k2 andn2 for the slow reacting groundwater storage, while variables k1and n1 for the fast reacting groundwater storage are fixed in eachrow. In the upper row for k1 is equal to 1.75 and n1 = 4, in the sec-ond row k1 = 1.6 and n1 = 4. Each column represents simulationsfor different c, which divides the groundwater recharge to thetwo-groundwater storages. Fig. 8 shows that a parameter

n strategy and non-calibrated a priori model.

Fig. 7. Observed vs. simulated spring discharges for Ein el Asal headwater catchment for different parameter sets of k and n using a model structure of only one subsurfacereservoir. The 1:1 Line is given in grey.

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 201

combination of c = 0.7, k1 = 1.75, n1 = 4, k2 = 45 and n2 = 2 resultsin the highest correlation. Although the correlation is not muchhigher than in the case of the single source continuum, the bias sig-nificantly decreased and simulation does not tend to underesti-mate observation anymore.

In Fig. 9 the seasonal variation of observed spring discharges inEin el Asal and the simulated baseflow are given. The solid blackline represents the observed discharge, while the dotted lines rep-resent simulation using a priori parameter values, with a varyingallocation of water to the two-groundwater storages. The simu-lated baseflow using a priori values does not reproduce the ob-served seasonal regime in Ein el Asal. The dashed and grey linewith circles represent the simulations using only one fast ground-water storage. Both simulations represent much better the sea-sonal regime, but tend to underestimate the baseflow. The solidgrey line represents the simulation of suing two groundwater sto-rages calibrated to spring discharges. Introducing a second ground-water storage result in higher baseflow estimates in the level of theobservations and the seasonal regime is nicely reproduced.

(1) Observations of flash flood runoff mirror the surface runoffdynamics. In the J2000 g model surface runoff dynamicsare controlled mainly by two calibration parameters: (a) bwhich controls potential evaporation and hence surpluswater, available for groundwater recharge and direct runoffand (b) the allocation parameter LVB, which allocates thesurplus water to the direct runoff component. Hence flashflood observations are used to calibrate the parameters b

and LVD. In Fig. 10 the observed and simulated dischargesfor flash flood events in the period 2000–2005 are givenfor different combinations of b and LVB. The few observedsurface runoff events typically react on the heavy precipita-tion events by a quick response and with a high flow rate,predominantly consisting of direct surface runoff. Anincrease of b causes higher potential evapotranspirationresulting in a decrease of runoff generation. Increasing LVDleads to more surface runoff components. The combinationLVD = 0.7, b = 1.2 and LVD = 0.8, b = 1.4 produce the best cor-relation between simulated and observed direct runoff vol-ume and timing.

(2) Groundwater recharge estimates derived from the CMB areused to estimate the remaining calibration parameters. Incontrast to the first two calibration steps, the parametersn, k, c, LVD and b, already addressed in calibration, wereallowed to be changed if necessary due to the following rea-sons. Observed spring discharges in headwaters are used tocalibrate parameters controlling groundwater recharge andit is assumed that the so calibrated parameters can be trans-ferred to the entire catchment. Groundwater recharge esti-mates for the entire catchment from CMB is used to testthe validity of this transfer. Observed flash floods are usedto calibrate surface runoff parameters LVD and b, that allo-cates water to surface runoff and groundwater recharge.The coefficient b also affects infiltration and hence theCMB is used to check the effect of b on groundwaterrecharge.

Fig. 8. Observed vs. simulated spring discharges for Ein el Asal headwater catchment for different parameter sets of k, n and c using a model structure with two subsurfacereservoirs. The 1:1 Line is given in grey.

Fig. 9. Observed spring discharge of Ein el Asal and simulated baseflow using two apriori parameter sets, two calibrated parameters sets for an model structure withonly one subsurface reservoir and a calibrated parameter set for an model structurewith two subsurface reservoirs.

202 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

To compare modelled groundwater recharge to the results ofthe CMB method, the modelling domain is extended to the subsur-face groundwater catchment (Fig. 1). The parameter combinationLVD = 0.8, b = 1.4 results in groundwater recharge estimate of43 mm/yr, while the parameter combination LVD = 0.7, b = 1.2 re-sults in an estimate of 61 mm/yr. The latter is very close to the esti-mates of 63 mm/yr derived from the chloride mass balance. Giventhe fact that groundwater recharge from CMB is, of course,

attributed to some uncertainty, which is much larger than the dif-ference to the modelled groundwater recharge, a further optimisa-tion of model parameters seem not to be reasonable.

Hence the nested multi-response calibration approach resultedin the following parameter set FCA = 1, ETR = 1, b = 1.2, LVD = 0.7,c = 0.7, k1 = 1.75, n1 = 4, k2 = 45 and n2 = 2.

4. Model validation

To validate the model the simulate groundwater recharge iscompared to estimates from the water table fluctuation (WTF)method. The results are compared to a model run with a prioriparameter values from literature as a benchmark.

In Fig. 11 the mean annual groundwater recharge rates for theperiod 1983–1997 from the WTF method, the modelling resultsusing the calibrated parameter set and the model results usingthe a priori parameter set as benchmark are given.

In the period 1989–1996 the groundwater recharge rates fromthe nested multi-response calibration show very similar valuesas the WTF model. However, in previous years the differences be-tween model and WTF method are larger. Particular in the period1983–1986, there is a remarkable discrepancy between outcomesof the J2000 g model and the WTF-method.

A reason could be uncertainties in rainfall. For that period norainfall observation in the south and north-east of the catchmentare available. Missing rainfall observations, particular in the south,hinders to correctly interpolate the spatial patterns of catchment

Fig. 10. Simulation of flash flood events using different parameter sets.

Fig. 11. Annual groundwater recharge rates derived from the water table fluctuation (WTF) method and the model simulation using calibrated parameters of the multi-response approach and uncalibrated a priori model parameters. Mean annual precipitation is given in grey and their harmonic mean is given as dashed grey line.

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 203

rainfall with its strong gradient from south to north and the rainfallas input into the hydrological model is expected to be underesti-mated. It becomes apparent by comparing annual precipitation

amounts (grey columns) and their harmonic mean (dashed line)in Fig. 11, the rainfall in the years 1983–1986 was clearly belowaverage rainfall rates.

Fig. 12. Spatial patterns of the regionalized mean annual precipitation (A) and groundwater recharge (B).

204 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

Generally, the model using the nested multi-response calibra-tion approach tends to better match the rates of the WTF methodthen a model with a priori values. The a priori model tends to over-estimate groundwater recharge rates. A strong overestimation ofrecharge rates by the a priori model is visible in the wet winterperiod of 1991/1992. The WTF model and the calibrated modelestimate recharge rates up to 230 mm/yr, while the a priori modelestimates recharge rates up to 328 mm/yr. This is an overestima-tion of about 30%.

5. Modelling results

After calibrating and validating, the model is used to simulategroundwater recharge for the period 1980–2008. In Fig. 12 the spa-tial patterns of mean annual rainfall (left) and groundwater re-charge (right) are given.

Fig. 13. Time series of groundwater recharge (A) and flash flood ru

The spatial patterns of annual precipitation (Fig. 12A) repre-sents the gradient from S and SE (Ajloun) well with annualamounts of between 450 and 565 mm. A similar spatial patternis visible for the groundwater recharge (Fig. 12B). High groundwa-ter recharge values of >100 mm/yr are connected to the mountain-ous parts of the catchment, while it decreases to less than 50 mm/yr in the region of the Lower Jordan Valley. However, recharge de-clines locally as well, e.g. along the flanks of the deeply incised wa-dis with steep slopes. Contrastingly, the base of the valleys ischaracterised by flatting and consequently higher groundwaterrecharge.

In Fig. 13 the time series of precipitation and groundwater re-charge (A – top) and flash flood runoff (B – bottom) simulated bythe J2000 g model are given for the period 1980–2008. The modelnicely simulates the seasonal patterns of groundwater rechargeand flash flood runoff in the wet winter period. The modelling

noff (B) for the Wadi Al Arab simulated by the J2000 g model.

T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206 205

results clearly visualize the strong connection of rainfall dynamicsand ground water recharge and flash flood runoff. It indicates thatin seasons in which monthly rainfall rates do not exceed values of100 mm/month, no ground recharge takes place. To generate flashflood runoff, monthly rainfall rates of more than 115 mm/month inthe rainy season are necessary.

Using the model and observed rainfall the averaged water bud-get components are estimated to amount to 489 mm/yr of precip-itation, 413 mm/yr of actual evaporation, 15 mm/yr of direct runoff(flash flood) and 61 mm/yr of groundwater recharge.

6. Conclusions

In the paper a nested multi-response calibration approach for asemi-distributed hydrological model is developed in the semi-aridAl-Arab catchment, for which runoff data is only sparsely available.

First the available different data source were screened for infor-mation content of processes, e.g. if data sources contain informa-tion on mean values, spatial or temporal variability etc. for thecatchment or only sub-catchments. In a second step, the informa-tion content was mapped to the relevant model components,which represent these processes. Then the data source was usedto calibrate the respective subset of model parameters, while theremaining model parameters remained unchanged. This mappingis repeated for other available data sources.

Similar to the work of Sophocleous (1991) and Sharda et al.(2006) different data sets were used to confine model parametersand to reduce parameter uncertainty. However, most other studiesthat used more than one data source for model calibration cali-brated all model parameters at once (e.g. Mroczkowski et al.,1997; Western and Grayson; 2000; Parajka et al., 2007a,b; Marceet al., 2008) using a multi-objective optimization function. Due tovery limited information content of the available data source, i.e.long term mean values of groundwater recharge vs. time seriesof soil moisture, parameters are calibrated in a consecutive man-ner. This limits the often-visible effect that unrealistic values ofone parameter can be compensated by other parameters. The con-secutive calibration is based on the assumption that a distinctmapping of single model parameters to single hydrological pro-cesses is possible. However, many parameters in conceptual modeleffect more than one process, e.g. allocation parameters whichallocate water in the root zone to surface runoff or groundwater re-charge clearly affect both processes, and calibrating such parame-ters only to one process may result in sub-optimal parameterranges. We believe that in catchments with data of very limitedinformation content, i.e. the absence of time series of meaningfullength and resolution, a consecutive calibration tends to be in morereliable parameter range, than calibrating all parameters at oncerisking compensating unrealistic parameters by other parameters.

In that study the gauged spring discharge (GSD) method, flashflood observations and data from the chloride mass balance(CMB) are used to derive plausible parameter ranges. The water ta-ble fluctuation (WTF) method is used to validate the model. Resultsfrom modelling using a priori parameter values from literature as abenchmark are compared. The performance of calibrated model invalidation is better than the model using a priori parameters val-ues. The estimated recharge rates of the calibrated model are smal-ler the ±10% of the estimates derived from WTF method. Largerdifferences between model estimates and WTF estimates are visi-ble in the years 1983/1986, but are attributed to uncertainties inthe spatial patterns of rainfall as input into the hydrological model.This underlines the strong effects of input data on the model out-put. It is quite clear that a hydrological model can only provide reli-able estimates of hydrological quantities, if uncertainties inobserved rainfall are not too high.

The model with a priori parameter values tends to overestimaterecharge rates, particular in the wet winter period of 1991/1992,during which recharge rates were overestimated by up to 30%. Thisin turn would lead to erroneous estimations of available water re-sources and clearly endangers reliable water resource managing inwater scarce regions.

There are still large uncertainties in the estimated spatio-tem-poral patterns of groundwater recharge derived from the modelusing the nested mulit-response calibration approach. This is dueto the large uncertainties in input data such as rainfall and uncer-tainty and limited information content in calibration data. How-ever, Kinzelbach et al., 2002 clearly stated that estimation ofgroundwater recharge in semi-arid and arid region is very complexand is generally affected by uncertainties. The proposed nestedmulti-response approach may help to better predict water re-sources despite data scarcity by exploiting all availableinformation.

Acknowledgements

The authors thank the German Federal Ministry of Educationand Research for funding the SMART project (funding code: 02-WM0802). The authors particularly thank Secretary General Assis-tant Eng. Ali Subah, the Ministry of Water and Irrigation Jordan, Dr.K. Hadidi, the Water Authority of Jordan, Jordan Valley Authorityand Dr. Armin Margane from the Federal Institute for Geosciencesand Natural Resources - BGR for fruitful cooperation and the kindprovision of data.

References

Abdulla, F., Al-Assa’d, T., 2006. Modeling of groundwater flow for Mujib aquifer,Jordan. J. Earth Syst. Sci. 115 (3), 289–297.

Acherar, M., Rambal, S., 1992. Comparative water relations of four Mediterraneanoak species. Plant Ecol. 99–100, 177–184.

Al-Abed, N., Abdullah, F., Abu Khyarah, A., 2005. GIS-hydrological models formanaging water resources in the Zarqa River basin. Environ. Geol. 47, 405–411.

AL-Eisawi, D., 1996. Vegetation of Jordan. UNESCO, Cairo Office, pp. 29–43.Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspiration,

Guidelines For Computing Crop Water Requirements. FAO Irrig. and Drain.Paper 56, Food and Agric. Orgn. of the United Nations, Rome, Italy. 300 pp.

Amraoui, F., Razack, M., Bouchaou, L., 2013. Turbidity dynamics in karstic systems.Example of Ribaa and Bittit springs in the Middle Atlas (Morocco). Hydrol. Sci. J.48 (6), 971–984.

Bajjali, W., 2006. Recharge mechanism and hydrochemistry evaluation ofgroundwater in the Nuaimeh area, Jordan, using environmental isotopetechniques. Hydrogeol. J. 14, 180–191.

Baldocchi, D.D., Xu, L., Kiang, N.Y., 2004. How plant functional-type, weather,seasonal drought, and soil physical properties alter water and energy fluxes ofan oak-grass savanna and an annual grassland. Agr. Forest Meteorol. 123, 13–39.

Bazuhair, A.S., Wood, W.W., 1996. Chloride mass-balance method for estimatingground water recharge in arid areas: example from western Saudi Arabia. J.Hydrol. 186, 153–159.

Berndtsson, R., Larson, M., 1987. Spatial variability of infiltration in a semi-aridenvironment. J. Hydrol. 90, 117–133.

Beven, K.J., Binley, A.M., 1992. The future of distributed models – model calibrationand uncertainty prediction. Hydrol. Process. 3, 279–298.

BGR (Deutsche Bundesanstalt für Geowissenschaften und Rohstoffe), WAJ (WaterAuthority of Jordan), 1995. Groundwater Resources of Northern Jordan,Groundwater Abstraction, Groundwater Monitoring, vol. 2, part 1. Ministry ofWater and Irrigation, Amman, Jordan.

BGR (Deutsche Bundesanstalt für Geowissenschaften und Rohstoffe), WAJ (WaterAuthority of Jordan, 1996. Groundwater Resources of Northern Jordan, Rainfall,Spring Discharge And Baseflow, vol. 1, part 1. Ministry of Water and Irrigation,Amman, Jordan.

BGR (Deutsche Bundesanstalt für Geowissenschaften und Rohstoffe), WAJ (WaterAuthority of Jordan, 2001. Groundwater Resources of Northern Jordan –Contributions to the Hydrology of Northern Jordan, vol. 4. Ministry of Waterand Irrigation, Amman, Jordan.

Blöschl, G., Grayson, R., 2002. Advances in Distributed Hydrological Modelling—Towards a New Paradigm. In: Third International Conference on WaterResources and Environmental Research, Dresden.

Dorman, J.L., Sellers, P.J., 1989. A global climatology of albedo, roughness length andstomata] resistance for atmospheric general circulation models as representedby the Simple Biosphere Model. J. Appl. Meteorol. 28, 833–855.

206 T. Rödiger et al. / Journal of Hydrology 509 (2014) 193–206

Gräbe, A., Rödiger, T., Rink, K., Fischer, T., Sun, F., Wang, W., Siebert, C., Kolditz, O.,2012. Numerical analysis of the groundwater regime in the western Dead Seaescarpment, Israel + West Bank. Environ. Earth Sci.. http://dx.doi.org/10.1007/s12665-012-1795-8.

Healy, R.W., Cook, P.G., 2002. Using ground-water levels to estimate recharge.Hydrogeol. J. 10 (1), 91–109.

IAEA/WMO, 2006. Web Site of the Global Network of Isotopes in Precipitation(GNIP) and Isotope Hydrology Information System (ISOHIS). <http://isohis.iaea.org/>, IAEA, Vienna.

Kattan, Z., 2006. Characterization of surface water and groundwater in theDamascus Ghotta basin: hydrochemical and environmental isotopesapproaches. Environ. Geol. 51, 173–201.

Kergoat, L., 1998. A model for hydrological equilibrium of Leaf Area Index on aglobal scale. J. Hydrol. 212–213, 268–286.

Kinzelbach, W., Aeschbach, W., Alberich, C., Goni, I.B., Beyerle, U., Brunner, P.,Chiang, W.H., Rueedi, J. and Zoellmann, K., 2002. Asurvey of Methods forGroundwater Recharge in Arid and Semi-Arid Regions. Early Warning andAssessment Report Series, UNEP/DEWA/RS.02-2. United Nations EnvironmentProgramme, Nairobi, Kenya.

Klemeš, V., 1986. Operational testing of hydrological simulation models. Hydrol. Sci.J. 31, 13–24.

Körner, Ch., 1994. Leaf diffusive conductances in the major vegetation types of theglobe. In: Schulze, E.D., Goldwell, M.M. (Eds.), Ecophysiology of Photosynthesis,Ecological Studies, vol. 100. Springer, Berlin, pp. 463–490.

Kralisch, S., Krause, P., 2006. JAMS – A Framework for Natural Resource ModelDevelopment and Application. In: Gourbesville, P., Cunge, J., Guinot, V., Liong,S.-Y. (Eds.), Proceedings of the 7th International Conference onHydroinformatics, pp. 2356–2363.

Krause, P., Hanisch, S., 2009. Simulation and analysis of the impact of projectedclimate change on the spatially distributed waterbalance in Thuringia,Germany. Adv. Geosci. 21, 33–48.

Krause, P., Biskop, S., Helmschrot, J., Flügel, W.-A., Kang, S., Gao, T., 2010.Hydrological system analysis and modelling of the Nam Co Basin in Tibet.Adv. Geosci. 27, 29–36.

Madsen, H., 2003. Parameter estimation in distributed hydrological catchmentmodeling using auto- automatic calibration with multiple objectives. Adv.Water Resour. 26, 205–216.

Marce, R., Ruiz, C.E., Armengol, J., 2008. Using spatially distributed parameters andmulti-response objective functions to solve parameterization of complexapplications of semi-distributed hydrological models. Water Resour. Res. 44(2), W02436. http://dx.doi.org/10.1029/2006WR005785.

Ministry of Agriculture, Jordan, 1994. National Soil Map and Land Use Project – TheSoils of Jordan, Level 2: Semi-detailed Studies, vol. 2. Main Report, Ministry ofAgriculture, Jordan.

Mroczkowski, M., Raper, G.P., Kuczera, G., 1997. The quest for more powerfulvalidation of conceptual catchment models. Water Resour. Res. 33, 2325–2336.

Nash, J.E., 1958. The form of the instantaneous unit hydrograph. Int. Assoc. Sci.Hydrol., Publ. n62 3, pp. 114–118.

Niinemets, Ü., Cescatti, A., Rodeghiero, M., Tosens, T., 2005. Leaf internal diffusionconductance limits photosynthesis more strongly in older leaves ofMediterranean evergreen broad- leaved species. Plant, Cell Environ. 28, 1552–1566.

Parajka, J., Blöschl, G., Merz, R., 2007a. Regional calibration of catchment models –potential for ungauged catchments. Water Resour. Res.. http://dx.doi.org/10.1029/2006WR005271.

Parajka, J., Merz, R., Blöschl, G., 2007b. Uncertainty and multiple objectivecalibration in regional water balance modelling. Hydrol. Process.. http://dx.doi.org/10.1002/hyp.6253.

Rhizopoulou, S., Mitrakos, K., 1990. Water relations of evergreen sclerophylls I.Seasonal changes in the water relations of eleven species from the sameenvironment. Ann. Bot. 65, 171–178.

Rödiger, T., Sauter, M., Büchel, G., 2009. Infiltration und Grundwasserströmung ingeklüftet-porösen Buntsandsteingrundwasserleitern im Osten des ThüringerBeckens. Grundwasser – Zeitschrift der Fachsektion Hydrogeologie 14, 21–32.

Schulze, E.D., Kelliher, F.M., Körner, C., Lloyd, J., Leuning, R., 1994. Relationshipbetween maximum stomatal conductance, ecosystem surface conductance,carbon assimilation rate and plant nitrogen nutrition: a global ecology scalingexercise. Ann. Rev. Ecol. Syst. 25, 629–660.

Sharda, V.N., Kurothe, R.S., Sena, D.R., Pande, V.C., Tiwari, S.P., 2006. Estimation ofgroundwater recharge from water storage structures in a semi-arid climate ofIndia. J. Hydrol. 329, 224–243.

Sophocleous, M., 1991. Combining the soilwater balance and water-levelfluctuation methods to estimate natural groundwater recharge: practicalaspects. J. Hydrol., 229–241.

Tilch, N., Uhlenbrook, S., Leibundgut, Ch., 2002. Regionalisierungsverfahren zurAusweisung von Hydrotopen in von periglazialem Hangschutt geprägtenGebieten. Grundwasser – Zeitschrift der Fachsektion Hydrogeologie 4 (2002),206–216.

USDA, 1975. Soil Taxonomy, Agricultural Handbook No. 436, United State,Department of Agriculture.

Western, A., Grayson, R., 2000. Soil Moisture and Runoff Processes at TarrawarraChapter 9. In: Grayson, R., Blöschl, G. (Eds.), Spatial Patterns in CatchmentHydrology: Observations and Modelling. Cambridge University Press,Cambridge, UK, pp. 209–246.

Wilson, M.F., Henderson-Sellers, A., 1985. A global archive of land cover and soilsdata for use in general circulation climate models. J. Clim. 5, 119–143.

Wood, W., Sanford, W., 1995. Chemical and isotopic methods for quantifyinggroundwater recharge in a regional, semiarid environment. Ground Water 33,458–468.

Wu, Y., Wang, W., Toll, M., Alkhoury, W., Sauter, M., Kolditz, O., 2011. Developmentof a 3D groundwater model based on scarce data: the Wadi Kafrein catchment/Jordan. Environ. Earth Sci. 64 (3), 771–785.

Zohary, M., 1973. Geobotanical Foundations of the Middle East, vol. 2. GustavFischer Verlag, Germany.