Rainfall-Runoff Modeling in Arid Areas

By the Faculty for Geosciences, Geotechnique and Mining of the Technische Universitt Bergakademie Freiberg approved

THESIS

to attain the academic degree of Doctor rerum naturalium Dr. rer. nat

Submitted by M.Sc. Eng. Eyad Hamad Abushandi Born on the April 14th, 1980, in Irbid, Jordan

Assessors: Prof. Dr. Broder Merkel PD Dr. Volkmar Dunger

Date of the award: 08.04.2011, in Freiberg

Preface
The present PhD dissertation Rainfall-Runoff Modeling in Arid Areas, Jordan has been submitted as a part of the requirement for the Ph.D. degree at the Technical University of Freiberg (TUBAF). The PhD supervisor was Professor Dr. Broder Merkel. The research was partially funded by Schsisches Landesstipendium, DAAD, and BAFG-Amt. The thesis is organized as a synopsis and an appendix, of which the synopsis contains an introduction, the objectives, review of existing work, and a short outline of three manuscripts in international journals with peer review in place. The study has taken place at the Department of Geology from October 2007 to March 2011. An external research stay of two months was spent at the Environment Centre, The Lancaster University/ England, with Professor Keith Beven. Three field campaigns were essential for developing cooperation with local partner universities: University of Jordan, and Al-Albyat University, gathering data of geo-referenced objects as well as from the Jordanian water authorities, and installing a stand alone weather station.

Freiberg, March 2011, MSc. Eng. Eyad Hamad Abushandi

This thesis is dedicated to my mother, my wife, and to the soul of my father

II

Acknowledgements
I gratefully acknowledge my supervisor Prof. Dr. Broder J. Merkel for his inspiration, competent guidance, patience, and encouragement through all the different stages of this research. I really thank him for guiding me to conduct my PhD research in a challenging and enjoyable topic of hydrological modeling application in arid regions. A special thank goes to PD Dr. Volkmar Dunger, especially for his valuable comments and discussions to improve the research quality. My thanks also go to all colleagues of the hydrogeology team at the Geology Institute for many good discussions and for making my stay at TUBAF enjoyable: Sascha Kummer, Mandy Schipek, Robert Sieland, Hajo Peter, Mohammed Khattab, Iwona Woloszyn, Nadja Schmidt, Ramadan Abdel Aziz, Zheenbek Kulenbekov, Lotfallah Karimzadeh, Nair Sreejesh, Sameh Wisam, Wondem Gezahegne, Yvonne Lindig, Andrea Berger, Ali Alaoui, Samer Bachmaf, and Rudy Abo. A special thank to my officemates Sosina Shieleses and Raghid Sabri for creating a very nice atmosphere. My thanks also go to our technical and administration staff members of the department for their help during my PhD especially Tino Mai, Maja Merz, and Gisela Schmiedgen. I am really grateful to the university administration office especially Dagmer Heim, Corina Dunger, from the BAFGAmt Annette Kunze. My deep thank for all Jordanian partners for the fruitful collaboration and data exchange, especially Prof Dr. Wajih Owais from the Jordan University of Science and Technology and his help of writing official letters to local Jordanian authorities and to Al-Albayt University for offering a suitable placement to install the weather station. To make the two months external research in the UK possible I would like to thank University of Lancaster, Professor Keith Beven and furthermore the financial support from the BAFG-Amt in Freiberg. I am forever indebted to my mother for her understanding, endless patience and encouragement when it was most required throughout the study-period. Thanks to my brothers and sisters specially Khalid, Ayman, Raghda, and Mai. I am deeply indebted for not III

having spent much time with them throughout my study in Germany. Their love reminded me of what is truly important and helped me maintain perspective on my life while completing this research. I would like to thank my father Burhan Al-Qasem, my mother in low Ameera Al-Naser, to my aunt Lamia Al-Qasem for their patience and support. Last, but not least, this thesis would not have been possible without the complete support of my wife Hala Al-Qasem.

IV

Abstract
The Wadi Dhuliel catchment/ North east Jordan, as any other arid area has distinctive hydrological features with limited water resources. The hydrological regime is characterized by high variability of temporal and spatial rainfall distributions, flash floods, absence of base flow, and high rates of evapotranspiration. The aim of this Ph.D. thesis was to apply lumped and distributed models to simulate stream flow in the Wadi Dhuliel arid catchment. Intensive research was done to estimate the spatial and temporal rainfall distributions using remote sensing. Because most rainfall-runoff models were undertaken for other climatic zones, an attempt was made to study limitations and challenges and improve rainfall-runoff modeling in arid areas in general and for the Wadi Dhuliel in particular. The thesis is divided into three hierarchically ordered research topics. In the first part and research paper, the metric conceptual IHACRES model was applied to daily and storm events time scales, including data from 19 runoff events during the period 1986-1992. The IHACRES model was extended for snowfall in order to cope with such extreme events. The performance of the IHACRES model on daily data was rather poor while the performance on the storm events scale shows a good agreement between observed and simulated streamflow. The modeled outputs were expected to be sensitive when the observed flood was relatively small. The optimum parameter values were influenced by the length of a time series used for calibration and event specific changes. In the second research paper, the Global Satellite Mapping of Precipitation (GSMaP_MVK+) dataset was used to evaluate the precipitation rates over the Wadi Dhuliel arid catchment for the period from January 2003 to March 2008. Due to the scarcity of the ground rain gauge network, the detailed structure of the rainfall distribution was inadequate, so an independent from interpolation techniques was used. Three meteorological stations and six rain gauges were used to adjust and compare with GSMaP_MVK+ estimates. Comparisons between GSMaP_MVK+ measurements and ground rain gauge records show distinct regions of V

correlation, as well as areas where GSMaP_MVK+ systematically over- and underestimated ground rain gauge records. A multiple linear regression (MLR) model was used to derive the relationship between rainfall and GSMaP_MVK+ in conjunction with temperature, relative humidity, and wind speed. The MLR equations were defined for the three meteorological stations. The best fit of the MLR model for each station was chosen and used to interpolate a multiscale temporal and spatial distribution. Results show that the rainfall distribution over the Wadi Dhuliel is characterized by clear west-east and north-south gradients. Estimates from the monthly MLR model were more reliable than estimates obtained using daily data. The adjusted GSMaP_MVK+ dataset performed well in capturing the spatial patterns of the rainfall at monthly and annual time scales, while daily estimation showed some weakness for light and moderate storms. In the third research paper, the HEC-HMS and IHACRES rainfall runoff models were applied to simulate a single streamflow event in the Wadi Dhuliel catchment that occurred in 3031.01.2008. Both models are considered suitable for arid conditions. The HEC-HMS model application was done in conjunction with the HEC-GeoHMS extension in ArcView 3.3. Streamflow estimation was performed on hourly data. The aim of this study was to develop a new framework of rainfall-runoff model applications in arid catchment by integrating a readjusted satellite derived rainfall dataset (GSMaP_MVK+) to determine the location of the rainfall storm. Each model has its own input data sets. HEC-HMS input data include soil type, land use/land cover map, and slope map. IHACRES input data sets include hourly rainfall and temperature. The model was calibrated and validated using observed stream flow data collected from Al-Zaatari discharge station. IHACRES shows some weaknesses, while the flow comparison between the calibrated streamflow results agrees well with the observed streamflow data of the HEC-HMS model. The Nash-Sutcliffe efficiency (Ef) for both models was 0.51, and 0.88 respectively. The application of HEC-HMS model in this study is considered to be satisfactory.

VI

Kurzfassung (German)
Das Untersuchungsgebiet Wadi Dhuliel im Nordosten Jordaniens hat wie alle anderen ariden Gebiete besondere hydrologische Eigenschaften mit begrenzten Wasserressourcen. Das hydrologische System ist durch hohe Variabilitt der zeitlichen und rumlichen Verteilungen von Niederschlag, Durchfluss, des Fehlens einer Basisabflusskomponente und hohe Verdunstung charakterisiert. Das Hauptziel dieser Doktorarbeit ist Anwendung verschiedener Niederschlags-Abflussmodells (Lumped und distributed ) um das Abflussverhalten im Einzugsgebiet des Wadu Dhuliel zu simulieren. Diese Modelle sind wichtig, um rumliche Zusammenhnge und Rckkopplungsmechanismen verschiedener hydrologischer Variablen in ariden Gebieten zu verstehen und die tatschliche Wasserbilanz zu beschreiben. Dies erlaubt einen realistischen berblick ber alle Aspekte der Nachhaltigkeit. Die Dissertation ist in eine Synopse und einen Anhang eingeteilt. Die Synopse enthlt eine Einleitung, die Ziele, die Literaturrecherche und einen kurzen berblick ber drei Publikationen. Die Anlage umfasst drei Publikationen, welche die eigentliche wissenschaftliche Arbeit dieser Dissertation bilden. Das Untersuchungsgebiet (Wadi Dhuliel) ist vor allem Ackerland; es befindet sich im Nordosten Jordaniens mit einer Flche von 1985 km2. Die Wahl Gebietes wurde auf der Grundlage der Kombinationen von kologischen und anthropogenen Faktoren getroffen, die Einfluss auf die hydrologische Bilanz in diesem Gebiet Jordaniens haben. Darber hinaus ist das Untersuchungsgebiet Teil des Badia Forschungsund

Entwicklungsvorhabens. In der ersten Forschungsarbeit wurde das konzeptionelle Niederschlag-Abfluss-Modell IHACRES auf tgliche Zeitskalen und Starkregenereignisse unter Einbeziehung von Daten von 19 (Oberflchen)Abflussereignissen (Zeitraum 1986-1992) angewendet. Die

ursprngliche Struktur des Modells IHACRES verwendet einen exponentiell abklingenden Bodenfeuchte-Index, Infiltrationsrate und Evapotranspiration, um den Niederschlag in effektiven Niederschlag umzuwandeln. Eingabeparameter in IHACRES sind a) Niederschlag,

VII

b) Temperatur (optional) und c) Abflussmessungen ( zur Kalibrierung des Modells). Die zeitliche Auflsung erfolgt stndlich, tglich und monatlich ohne Lcken. Das IHACRESModell wurde fr Schneefall erweitert, um solche Extremereignisse mit einzubeziehen. Whrend das Ergebnis des IHACRES-Modells auf der Basis tglicher Werte mangelhaft ist, zeigte die Modellierung fr Starkniederschlagsereignisse eine gute bereinstimmung zwischen gemessenen und simulierten Durchflusswerten. Das Modell reagiert sehr sensitiv fr kleine Abflussereignisse. Die optimalen Parameterwerte wurden durch die Lnge der Kalibrierungszeitrume und ereignisspezifische nderungen beeinflusst. In der zweiten Forschungsarbeit wurde der Datensatz der globalen Satellitenkartierung von Niederschlgen (GSMaP_MVK+) verwendet, um die Niederschlagsraten des Einzugsgebietes Wadi Dhuliel fr den Zeitraum von Januar 2003 bis Mrz 2008 zu bewerten. Aufgrund der geringen Zahl von an Niederschlagssammlern konnten die detaillierte Struktur der Niederschlagsverbreitung nicht angemessen bestimmt werden; deshalb wurde eine unabhngige Interpolationstechnik verwendet. Drei meteorologische Stationen und sechs Regensammler wurden benutzt, um Schtzungen des GSMaP_MVK+ Datensatzes zu optimieren. Ein Vergleich von GSMaP_MVK+ Messungen und Bodenniederschlagsmessungen zeigt bestimmte Gebiete, die miteinander korrelieren, sowie Gebiete, in denen die GSMaP_MVK+ Daten systematisch die Bodenniederschlagsmessungen berund

unterschtzen. Ein multiples lineares Regressionsmodell (MLR) wurde angewendet, um die Beziehungen zwischen Niederschlag und dem GSMaP_MVK+ Datensatz, in Verbindung mit Temperatur, relativer Luftfeuchtigkeit und Windgeschwindigkeit abzuleiten. Die MLRGleichungen wurden fr die drei meteorologischen Stationen definiert. Die beste Anpassung des MLR-Modells fr jede Station wurde ausgewhlt und dazu verwendet, eine zeitliche und rumliche Verteilung zu interpolieren. Die Ergebnisse zeigen, dass die Verteilung der Niederschlge ber dem Wadi Dhuliel durch ein deutliches West-Ost- und Nord-Sd-Geflle geprgt ist. Schtzungen aus dem monatlichen MLR-Modell erbrachten bessere Korrelationen als Schtzungen, die sich aus der Verwendung von Tageswerten ergaben. Die angepassten VIII

GSMaP_MVK+ Daten eignen sich gut fr die Erfassung der rumlichen Muster von Niederschlgen in monatlichen und jhrlichen Zeitskalen, wohingegen tgliche Schtzungen Schwchen bei leichten und moderaten Starkregenereignissen zeigten. In der dritten Forschungsarbeit wurden die hydrologischen Modelle HEC-HMS und IHACRES angewendet, um ein einzelnes Abflussereignis zu simulieren, das vom 30.31.01.2008 im ariden Wadi Dhuliel-Einzugsgebiet auftrat. Beide Modelle werden als geeignet fr aride Bedingungen eingeschtzt. Die Anwendung des Modells HEC-HMS erfolgte in Zusammenarbeit mit der HEC-GeoHMS-Erweiterung in ArcView 3.3.

Oberflchenabflussschtzungen wurden auf der Grundlage von stndlichen Daten durchgefhrt. Das Ziel dieser Studie war es, durch die Integration von modifizierten, satellitengesttzten Daten der Niederschlagsmengen (GSMaP_MVK +) einen neuen Rahmen der Niederschlags-Abfluss-Modellanwendungen in ariden Einzugsgebiet zu schaffen, . Jedes Modell hat seinen eigenen Datensatz. Die HEC-HMS-Datenstze umfassen Bodenart, Bodennutzung/Bodenbedeckungkarte und eine Neigungskarte, die unter Verwendung von GIS enstanden. Die IHACRES-Eingangdatenstze umfassen stndliche Werte des Niederschlages und der Temperatur. Das Modell wurde kalibriert und validiert mit Hilfe beobachteter Durchflussdaten, die von der Messstation Al-Za'atari stammen. Das IHACRESModell zeigt einige Schwchen, wobei der Vergleich zwischen den kalibrierten Starkregendurchflssen und den beobachteten Durchflussdaten im HEC-HMS Modell eine gute bereinstimmung zeigt. Die Nash-Sutcliffe-Effizienz (Ef) fr beide Modelle betrgt jeweils 0,51 und 0,88. Die im Rahmen dieser Studie realisierte Anwendung des HEC-HMSModells wird als zufriedenstellend angesehen.

IX

Table of contents
Preface.........................................................................................................................................I Acknowledgements .................................................................................................................. III Abstract ..................................................................................................................................... V Kurzfassung (German) ............................................................................................................VII Table of contents ....................................................................................................................... X List of figures ..........................................................................................................................XII Introduction .............................................................................................................................. 14 1 2 3 3.1 3.2 3.3 3.4 4 4.1 4.2 5 6 7 Background and significance ........................................................................................... 14 Research objectives .......................................................................................................... 16 Study area description ...................................................................................................... 17 Rainfall ..................................................................................................................... 20 Temperature ............................................................................................................. 20 Surface runoff........................................................................................................... 21 Land use and soil types ............................................................................................ 24 Literature review .............................................................................................................. 25 Hydrological modeling in arid region ...................................................................... 25 Rainfall derived from remote sensing data .............................................................. 31 Deliverables...................................................................................................................... 32 Summary of appendices ................................................................................................... 34 Conclusions ...................................................................................................................... 35

References ................................................................................................................................ 37 Appendices: Appendix A Abushandi E. & Merkel B. 2010 Application of IHACRES rainfall-runoff model to the Wadi Dhuliel arid catchment, Jordan. J. Water and Climate Change (Published) ............................ 43 X

Appendix B Abushandi, E. and Merkel, B.: Rainfall estimation over the Wadi Dhuliel arid catchment, Jordan from GSMaP_MVK+, Hydrol. Earth Syst. Sci. Discuss., 8, 1665-1704, doi:10.5194/hessd-8-1665-2011, 2011. (Accepted for discussion).......................................... 74 Appendix C Abushandi E., Merkel B. (2011) Modelling rainfall runoff relations using HEC-HMS and IHACRES for a single rain event in arid region, Jordan. Water Resources Management Journal (Ready for submission).............................................................................................. 111

XI

List of figures
Figure 1. Average distribution of long-term (19382005) rainfall in Jordan. ......................... 15 Figure 2. Map of Al-Zarqa basin, including Seil Al-Zarqa subbasin and Wadi Dhuliel subbasin ......................................................................................................................................... 18 Figure 3. Al-Zarqa basin including the study area (Wadi Dhuliel) and the Seil Al-Zarqa subbasin ......................................................................................................................................... 19 Figure 4. Dry condition in the Wadi Dhuliel/ Northern Jordan .............................................. 22 Figure 5. Wadi Dhuliel, northern Jordan after a storm event .................................................. 22 Figure 6. Typical Stream flow event in the Wadi Dhuliel ....................................................... 23 Figure 7. Al-Zaatari gauging station including chart recorder ............................................... 23 Figure 8. Land use map of Wadi Dhuliel sub-basin................................................................. 24 Figure 9. General classification of hydrological models.......................................................... 27 Figure 10. Diver installation in the study area to measure streamflow magnitudes ................ 29 Figure 11. Integrating GIS with hydrological modeling .......................................................... 31

XII

Introduction 1 Background and significance

The scarcity of water and threats of flash floods in Jordan require further understanding the natural processes of water resources in order to manage and sustain current and future water resources. If current population and water consumption trends continue, by the year 2025, water supply per capita will fall from current 145m3/yr to 91m3/yr (Hadadin et al 2010). To make matters worse, rainfall fluctuation and climate change are expected to increase water scarcity in Jordan (Freiwan and Kadioglu 2008a). These combined factors will drive Jordan to severe water stress. Around 91% of Jordan lies in arid and semi-arid regions, which receive less than 200 mm of annual rainfall. The rainfall distribution displays a sharp westeast gradient, from relatively wet west regions with about 600mm per annum to dry east regions (the Jordanian desert) with rainfall less than 100mm per annum (Figure 1). Both the arid climate and topographic variations contribute to the variation in rainfall. However, only small fraction of the rainfall (8.5%) can be captured, while the rest is lost due to evapotranspiration (87.9%) and floods (3.7%) (JMWI). With rapid depletion of ground water resources, flood management could provide a potential means for futher optimize water resources. Since agriculture comprises much of the Jordanian economy, water availability is the vital factor controlling the economic growth in Jordan. Moreover, the management of infrequent flash floods requires detailed hydrological information about catchment.

14

Study area

Figure 1. Average distribution of long-term (19382005) rainfall in Jordan (Hadadin et al 2010), modified. The red square outlines the area studied within this thesis. Despite the importance of water resource management in Jordan, a substantial number of short-term flood flow studies have not yet been conducted for several reasons. (i) lack of awareness of the importance in monitoring floods (ii) cost of flood monitoring devices and (iii) difficulty of securely installing measurement devices within the study area. Even in the rare cases where gauging stations exist, measurement problems may occur during intense flood events, reducing the quality and completeness of the data (Lange et al 2000). The term aridity describes a dry hot climate, with a precipitation threshold less than 250 mm per year. Arid zones can be classified according to the degree of aridity.

15

de Martonne suggested the following expression for use as an aridity index (Lin 1999):

X = P /(10 + T ) where X is the aridity index, P is the annual precipitation in mm, and T is the mean annual air temperature in oC. Another aridity metric, the UNEP Aridity Index (Al) takes the ratio of average annual precipitation (P) to potential evapotranspiration (PET) (Safriel 2006) to

determine the degree of aridity. Based on the UNEP classification, the study area in this research is classified as an arid region.

Research objectives
The general objectives of this research are to: 1. collect high quality rainfall and runoff data for hydrological modeling, 2. assess the feasibility of lumped and distributed rainfall-runoff models in arid region, where monitored data are rare. 3. 4. critically analyze the selected models performance in the study area, develop a comprehensive analysis of spatial rainfall in the Wadi Dhuliel arid catchment using remote sensing data, 5. provide a platform on which future research can be developed using the available hydrological data. The following specific objectives are addressed in different appendices: 1. evaluate all available rainfall-runoff models in order to find two which perform adequately for the boundary conditions of a Jordanian arid catchment.

16

2. 3.

analyze the accuracy of the calibration method in that region. investigate rainfall characteristics of the Wadi Dhuliel catchment in north-eastern Jordan by comparing remote sense data with groundbased rain gauge data

4.

develop a technique to enhance remote sense data of rainfall variability in arid regions using interpolation techniques rain gauge data and standard

5.

reduce the potential errors of rain gauge estimates and produce improved catchment scale rainfall distribution maps

6.

examine the performance of different models during a single heavy rainstorm event that caused streamflow in the Wadi Dhuliel subbasin

7.

parameterize the Wadi Dhuleil terrain using a high resolution digital terrain model (ASTER)

Study area description

Wadi Dhuliel, located in northeast Jordan, fulfils two important water

management-related functions: the discharge of the Al-Zarqa River and natural groundwater recharge. Most of the catchment area belongs to the Al-Zarqa Basin. In Arabic, the term Wadi is defined as a channel of water that is dry except during heavy rain events. Around 10% of the upper part of Wadi Dhuliel catchment is situated on Syrian territory (Figures 1 and 2). The region has an arid climate with cold, rainy winters and a hot, dry summer.

17

Figure 2. Map of Al-Zarqa basin, including Seil Al-Zarqa subbasin and Wadi

Dhuliel sub-basin (map datum: WGS 84)

The Wadi Dhuliel catchment, a typical example of an arid catchment, passes floodwater from northern Jordan into the Al-Zarqa River. The area is characterized by a very gently undulating topography with elevation varying from 512 m in the southwest to 1400 m in the north. The area of interest is approximately 1985 km2 (Figure 3).

18

Figure 3. Al-Zarqa basin including the study area (Wadi Dhuliel) and the Seil

Al-Zarqa sub-basin Besides Wadi Dhuliel, there are a number of side-wadis in the catchment area such as Wadi Al-Zaatari, Wadi Al-Ajeb and Wadi Alghar. The flow direction is from the north towards the Seil Al-Zarqa subbasin in the southwest. Irrigation for agriculture in the Wadi Dhuliel taxes existing groundwater resources and strongly distorts hydrological features of this arid catchment. The following general arid region characteristics describe the Wadi Dhuliel arid catchment: (i) Stream flow characterized by absence of base flow and flash-floods during infrequent high-intensity rain events. (ii) (iii) High evaporation rates. Highly-localized spatial rainfall distribution (Wheater et al 1991)--the storm rainfall correlation coefficient decreases rapidly with distance (Osborn et al 1979). 19

(iv)

Sparse plant cover and organic matter (McIntyre and Al-Qurashi 2009; Pilgrim et al 1988).

3.1

Rainfall

Short, high-intensity rain storms account for the dominant contribution to the low annual total rainfall in arid regions. Rainfall in the Wadi Dhuliel tends to vary markedly from year to year with an irregular distribution in time and space. As an illustrative example of the extreme yearly variability in Wadi Dhuliel, one rain gauge measured the annual rainfall to be 275.7, 93.1, 111.1, 230.4, 194.8, 63.1, and 209.5 mm over seven years. On one single day, 62 mm of rainfall occurred, even though the total annual rainfall in the same year was 100 mm (Sukhnah rain gauge). These kinds of rainfall events can generate significant surface runoff, resulting in severe soil erosion. Weather behavior and topographical characteristics play important roles in this variation. Around 73% of the total annual rainfall (mean 123 mm) occurs during November, December, January, and February.
3.2 Temperature

The area of interest is characterized by warm and dry summers (May to September) and moderate cold and wet winters (October to April) (Table 1). The average annual temperature is 16.8 C (19762005).

20

Table 1. Temperature description for the period of 19762005 (Um-Aljimal

Meteorological Station)
Jan. Average (C) Average maximum (C) Average minimum (C) 7.4 12.7 2.1 Jul. Average (C) Average maximum (C) Average minimum (C) 24.7 33.1 16.32 Feb. 8.6 14.2 2.9 Aug. 24.6 33 16.31 Mar. 11.5 18 5.1 Sep. 23.1 31.3 14.9 Apr. 16.2 24 8.5 Oct. 19.5 27.2 11.73 May 20.4 28.7 12.0 Nov. 13.6 20.2 6.92 Jun. 22.9 31.6 14.1 Dec. 9.0 14.5 3.51

3.3

Surface runoff

The Wadi Dhuliel catchment is exemplified by ephemeral wadis, where a stream runs fully for a short period of time, usually during and after heavy rain events (Figures 4 and 5), and is dry most of the year. Flash floods events fill desert dams and may recharge groundwater resources. The complex relationship between rainfall and streamflow is influenced by many factors, such as catchment slope, land cover type and density, soil type and infiltration rate, and evapotranspiration. Moreover, the quality and quantity of streamflow are strongly affected by urbanization and agricultural activity.

21

Figure 4. Dry condition in the Wadi Dhuliel/ Northern Jordan (13.04.2008)

Figure 5. Wadi Dhuliel, northern Jordan after a storm event (14.11.2008)

Flash flood hydrographs are generally sharp with durations between three hours and three days following the rain storm. In general, the runoff coefficient in the study area was 2.3% on average. Some flood flow events show more than one peak during a single rainstorm (Figure 6).

22

Figure 6. Typical Stream flow event in the Wadi Dhuliel (January

30-31, 2008) The only flood records available from the study area are provided from the AlZaatari gauging station, which is located on the lower part of the catchment area (Figure 7).

Figure 7. Al-Zaatari gauging station including chart recorder

(Stevens Type A) 23

3.4

Land use and soil types

The regional ecology, typified by vegetation cover, soil characteristics, plant and anthropogenic densities, affects the infiltration characteristics and influences the storage coefficient and runoff behaviour. Images from the Landsat Thematic Mapper (TM) allow categorization of land use in the Wadi Dhuliel into six classes (Figure 8 and Table 2). Much of the upstream land in the Wadi Dhuliel (36.3% of the total) of is largely cultivated while bare soil and settlements typify downstream surface cover.

Figure 8. Land use map of Wadi Dhuliel sub-basin, Landsat Thematic Mapper

(TM) of the year 1987, 30 m resolution. Table 2 Land use features for the Wadi Dhuliel area derived from TM (1987)
Land use type (%) Settlement Arable land Bare soil Basalt cover Seasonal cultivation Water body

Contribution (%)

34.2

13.17

19.7

9.6

23.11

0.23

24

The physical properties of the catchment soil govern water flow potential. The soil within the study area is primarily classified as Aridisols and belong to loam or silty loam. Aridisols contain high concentrations of lime (Al-Qudah 2001) withlow gypsum and basalt content in the subsoil and parent materials. Soil crusting most often occurs due to the high silt content in this type of soils.

4
4.1

Literature review
Hydrological modeling in arid region

Hydrological models provide a simplified mathematical representation of the hydrological system and are designed to model surface flow and/or groundwater processes and comprise essential tools for controlling and managing water resources. Additionally, hydrological models may increase our understanding of streamflow frequency (Reed et al 2007). Hydrological models are often employed due to the limitations of hydrological measurement techniques (Beven 2001b). Hydrological studies are often aimed at establishing rainfall-runoff relationships (Shah et al 1996). Rainfall-runoff models cover a wide range objectives and applications. Due to harsh climatic conditions and the lack of high quality observations, flood simulation in arid environment, especially flash flooding scenarios, is one of the greatest challenges facing hydrological modelers today. Different types of hydrological models exist to represent hydrological processes at a wide range of climatic and time scales. Rainfall-runoff models can be categorized according to the model type. The most common models are data-driven. Within this category, the rainfall-runoff

models falls into three sub-categories: (i) metric, (ii) conceptual, and (iii) physicalbased models (Wheater et al 1993). The models vary in complexity, amount of

25

observational data required as input parameters, and performance for different catchment types. Knowledge-driven systems comprise the second category. These rainfallrunoff models fall into two sub-types: lumped and distributed models. Lumped models treat the catchment as single homogenous unit. Previous published lumped models include the IHACRES model (Jakeman and Hornberger 1993), AWBM (Boughton and Chiew 2007), GR4J (Perrin et al 2003), and NAM (DHI 1993). Lumped models require limited amounts of input data at the expense of model resolution; detailed spatial fluctuations can not be resolved by these models. At the opposite end of the spectrum, distributed models make predictions by discretizing the catchment into a large number of grid squares (Beven 2001b). Available grid models include USDA SWAT (Arnold et al 1998b), TOPMODEL (Beven et al 1995), and HEC-HMS (Scharffenberg and Fleming 2010). Lumped models can be applied into catchments of limited data. Distributed models allow detailed description of hydrological processes which may fluctuate in space and time. Additionally, distributed models have the opportunity for parameterization with respect to geo-referenced objects within the catchment. Therefore, a great amount of spatially-related hydrological and physical data sets need to be prepared in order to run the model. Figure 8 shows the general classification of hydrological models.

26

Figure 9. General classification of hydrological models (Chow et al 2005)

modified The choice of the model depends on the study objectives; several considerations when must be taken into account when selecting a rainfall-runoff model: 1 Availability and the quality of hydrological data, especially rainfall and runoff data, 2 Model structure and model ability of regionalization, 3 Catchment characteristics and flow homogeneity. Many researchers recognize that rainfall-runoff models cannot give a full picture of reality (Al-Qurashi et al 2008; Beven 1984; Bronstert 2004; McIntyre and Al-Qurashi 2009; Pilgrim et al 1988); the spatial and temporal fluctuations in arid regions only intensify this dilemma for hydrologists working in these regions. Flash floods in arid regions result from extreme,irregular rainfall events. These events occur when conditions, such as soil moisture , infiltration conditions, steep slope (Rodier and Roche 1978), and high rainfall intensity (Gheith and Sultan 2002), 27

are fulfilled. The extreme irregularity of yearly flood events reduces the effectiveness of flood prediction methods. Flow measurement is particularly difficult in arid zones for several reasons: (i) low population density (Pilgrim et al 1988; Rodier and Roche 1978) (ii) few driveable roads, especially during in rainy (Pilgrim et al 1988) and flooding seasons (iii) lack of suitable natural control sections in streams with movable beds, and high cost of artificial control (Pilgrim et al 1988) (iv) (v) harsh climatological and physical conditions (Pilgrim et al 1988) moving rocks and debris may damage instruments placed in the Wadi (Kilpatrick and Cobb 1985) The quick rise and fall of the water level during flash floods makes flow measurements difficult (Lin 1999). Measuring flash floods in such areas are generally conducted via the velocity-area method (Rodier and Roche 1978) (e.g. Tunisia and China, Lin 1999). This method depends on measuring the average velocity of flow and calculating the cross sectional area of the Wadi. However, under unstable Wadi beds, this traditional flow measurement technique is problematic. Furthermore, turbulence and moving rocks pose could damage instruments for tracers and/or current-meter techniques. Therefore, new techniques of stream flow for arid environments should be developed to meet the project budget and overcome these challenges. Although a new diver from the Schlumberger Water Services Company was installed in the Wadi Dhuliel to measure water level during streamflow events (Figure 10), the resulting data set is rather short.

28

Figure 10. Diver installation in the study area to measure streamflow

magnitudes

Only a limited number of rainfall-runoff models have been developed and modified for arid environments. Osborn and Hickok published the first scientific research on characterization of rainfall-runoff relationships in 1968. They analyzed the effect of rainfall variability in producing runoff for the Alamogordo Creek watershed in New Mexico. For areas the size of the Wadi Dhuliel catchment, a few potential available rainfall-runoff models demonstrated acceptable performance in arid regions: the IHACRES model in Australia and South Africa and Oman (Croke et al 2006; Dye and Croke 2003; McIntyre and Al-Qurashi 2009), The KINEROS2 model in Oman and India (Al-Qurashi et al 2008; Sharma and Murthy 1996), the RORB3 model in Australia (Kotwicki 1987), the Pitman model in South Africa (Hughes et al 1997), the AHYMO model in the New Mexico (Schoener 2010), the HEC-HMS model in the UAE, West Bank-Palestine (Al-Rawas and Valeo 2008; Schoener 2010; Shadeed and Almasri 2010; Sherif et al 2011), , and the SWAT model in the UAE (Al Mulla 2005). 29

Abdulla et al. (2002) developed and applied a simple single event watershed model to simulate and predict the surface runoff hydrograph in the western part of Iraq. Their model was based on the water balance equation and their results showed good agreement between simulated runoff hydrograph and observed data. In Saudi Arabia, Al-Turbak (1996) applied a geomorphoclimatic model, where rainfall intensity and duration are calculated from physically-based infiltration components, in three arid catchments. Their results showed that this model could successfully reproduce surface runoff hydrographs, at least for events in which detailed and accurate data were available. In Israel, Lange et al., (2000) studied the importance of single events in arid zone rainfall-runoff modeling. The study concluded that separate analysis of single events is crucial for the understanding of high magnitudes floods in arid regions.Finally, the Spatial Water Budget Model (SWBM) and HEC-HMS / HEC-GeoHMS extension model were applied for managing water resources in the Zarqa River basin. The study of Al-Abed and Abu Khyara (2005) showed satisfactory results from both models. In this Ph.D. research study, the IHACRES and HEC-HMS rainfall runoff models were selected for several reasons: (i) model availability and structure, (ii) data availability, and (iii) model applicability in arid catchments. The model IHACRES (Identification of unit Hydrograph And Component flows from Rainfall, Evaporation and Stream flow data) is a simple model, parametrically efficient, and statistically rigorous (Dye and Croke, 2003). The IHACRES input requirements are only precipitation and temperature, and streamflow for calibration purpose. The other model, HEC-HMS (Hydrologic Engineering Centers Hydrologic Modeling System), requires inputs for the basin model, meteorologic model, and control specifications

30

but then allows several different techniques to be used to model the rainfall-runoff process.
4.2 Rainfall derived from remote sensing data

Hydrologists are extending their models to incorporate new data from remote, satellite-based rainfall (Geographic Information System or GIS) estimates for arid regions. GIS-compliance has already been implemented in many physical models, such as SWAT, HEC-HMS, and WEPP, not as data preparation tools but as a procedural step (Figure 11).
Data [on various servers]

Software [user driven client] Product [hardcopy & digital] GIS Spatial Analysis Maps
Visualization & Display

Land Use

Hydrology

Graphs/ Charts

Soils

Data Managment

Model [tighty coupled]

Tables

Weather Output Files Elevation

Model [un-coupled]

Model
[loosely coupled]

Statistics Package

Figure 11. Integrating GIS with hydrological modeling (Matson et al 1995)

31

Satellite-based rainfall estimates could compensate for the lack of high quality observational rainfall data as an input for rainfall runoff models in arid regions. Recently, a number of global precipitation determination systems have been developed to meet scientific demand, such as PERSIANN (Sorooshian et al 2000), Global Precipitation Climatology Project (GPCP) (Adler et al 2000; Huffman et al 2001; Xie et al 2003), and Multi-satellite Precipitation Analysis (TMPA) (Huffman et al 2007). Global or near-global satellite datasets are important to identify temporal and spatial rainfall changes in arid regions. Global Satellite Mapping of Precipitation (GSMaP) technique employs passive microwave radiometer data to estimate rainfall rates with mm/h precision (Ushio et al 2009). GSMaP combines precipitation retrievals from polar-orbiting satellites and cloud motion vectors derived from infrared images.

Deliverables

generate a weather parameters data set (rainfall, temperature , relative humidity, barometric pressure, and soil moisture) with 10 minute time steps over a range of 2 years by installing a stand alone weather station within the campus of Al-albayt University,

improve IHACRES model by adding an option for snow precipitation develop a new methodology to re-adjust satellite-derived rainfall datasets (GSMaP_MVK+) using ground based rainfall data

optimize the use of acquired weather data from the study area for present and future studies by designing a database management system (DBMS)

develop a water resource management plan for the Wadi Dhuliel.

32

The above-mentioned deliverables will improve the understanding of hydrological processes not only in the Wadi Dhuliel, by also in regions with different climatic and topographic variations. At a more general level, it could lead to a better understanding of the correlation between all weather variables. Despite limitations from the availability of input data, the results of this study help establish criteria for simulating streamflow in arid regions at different time scales. Specifically, this research explains how hydrological models should be used to model streamflow in arid regions when there is a dearth of observational data. The choice and quality of input data, model section, model parameters, model modifications, and applicability of results to different arid catchments on different timescales are discussed. The main research question was: what are the best hydrological models that can be applied to assess stream flow in arid regions under the lack of observational data. More specific questions have been addressed by this PhD research study: a. Which data should be collected to achieve research objectives? b. What is the quality of available data from local water authorities? c. What are the optimal sets of model parameter values? d. How can the research modify the model to meet the catchment characteristics? e. How can the research improve the accuracy of spatial rainfall estimation in arid regions? f. Is there any significance of using different rainfall runoff models in assessing same event or period?

33

The first four questions were answered in the first paper; the fifth question was addressed in the second paper, and the last question was answered in the third research paper.

Summary of appendices
The above targets have been achieved via publication of three articles in peer-

reviewed journals. The first publication was already published, the second has been accepted as a discussion paper, and the third is currently under review. The first author in all 3 papers is Eyad Abushandi. His contribution was 90% while the coauthor (Prof. Dr. Broder Merkel) contributed to 10% to the papers (ideas, proofreading). In the first paper, the metric conceptual IHACRES model was applied to daily time scales for storm events and was also extended for snowfall in order to cope with such extreme events. The detailed descriptions of the model, including model structure, formulation, data preparation, parameter identification, model calibration and evaluation, etc. were presented. Additionally, a general literature review on the subject was also incorporated. In the second paper, the investigation of rainfall characteristics over the Wadi Dhuliel catchment was madeutilizing Global Satellite Mapping of Precipitation GSMaP data and ground based rain gauge data. This study aimed to develop a technique to adjust or re-calibrate the GSMaP data by means of ground-based weather variables data and standard interpolation techniques. The study began with a detailed history recounting hydrological study in Jordan, then moved to the concept of using GSMaP_MVK+ in both arid reasons in general and the Wadi Dhuliel specifically. A statistical re- process was presented by which GSMaP_MVK+ data could be modified to account for complex terrain is the effects of rainfall variation with altitude were 34

briefly described. Finally, the influential factors for the over- or underestimation of the GSMaP_MVK+ in arid regions were summarized. The third paper discussed the advantages and the disadvantages of using distributed and lumped models. The applicability of GIS to hydrological modeling was highlighted and particular attention was given to explain the benefits of georeferenced datasets. The HEC-HMS distributed model and IHACRES lumped model were applied to simulate a single event streamflow in the Wadi Dhuliel arid catchment occurring from March 30-31, 2008. The paper began with a brief discussion about the concepts of both models and how GIS was integrated into the HEC-HMS model. The mathematical background of the models was then presented. The models were calibrated and validated using observed stream flow data set collected from AlZaatari discharge station and results presented.

Conclusions
In this study, a distributed hydrological model (HEC-HMS) and a lumped

model (IHCRES) are applied to the Wadi Dhuliel arid catchment to simulate streamflow on different time scales. Both models are applicable in arid areas of Jordan, which are dominated by ephemeral streams and rapid responses to storm events. The results indicate that both models can adequately simulate streamflow in arid catchments on the time scale of a single storm event but overall accuracy depends on the length of the chosen time interval. This study concluded that longer calibration periods were needed in order to reduce the uncertainty in model parameters, as well as predict long-term climate change and anthropogenic impact on the study area. A major source of uncertainty for the hydrological modeling is the lack of spatial rainfall input, despite the fact that a new parameter (snow melt) was added to the IHACRES model. Therefore, the second 35

publication focused on the incorporation of GSMaP_MVK+ data to simulate rainfall distribution. Although the GSMaP_MVK+ data proved extremely valuable for hydrological modeling, the new, satellite-derived rainfall data requires additional improvements for modeling continuous hydrological variables in arid catchments. In the third paper, the HEC-HMS and IHACRES models were applied for estimating the streamflow volume for a single rainfall event. The HEC-HMS rainfallrunoff model included the effects of terrain using GIS satellite data, which proved advantageous for describing for the variety land-use patterns and soil types.. On the other hand, the less-complex IHACRES model required minimal input data. The GSMaP_MVK+ dataset was used to determine the rain storm location and to quantify the amount of the rain in each sub-basin. Estimated and observed streamflow volumes of the single event were close enough to assume the applicability of the HEC-HMS model approach for the region.

36

References
Abdulla FA, Amayreh JA, Hossain AH, 2002, Single event watershed model for simulating runoff hydrograph in desert regions. Water Resources Management 16:221-238. Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and E. J. Nelkin, 2000, Tropical rainfall distributions determined using TRMM combined with other satellite and rain gauge information: Journal of Applied Meteorology, v. 39, p. 20072023. Al-Abed, N., F. Abdulla, and A. Abu Khyarah, 2005, GIS-hydrological models for managing water resources in the Zarqa River basin: Environmental Geology, v. 47, p. 405-411. Al-Turbak AS, 1996, Geomorphoclimatic peak discharge model with a physically based infiltration component. Journal of hydrology 176:1-12. Al Mulla, M. M., 2005, Application of a hydrological model in a data-poor arid region catchment: a case study of Wadi Ham, United Arab Emirates, Cranfield University, cranfield. Al-Qudah B, 2001, Soils of Jordan. In Zdruli S, Montanarealla L (eds), Soil Resources of Southern and Eastern Mediterranean Countries: CIHEAMIAMB, pp 127-141. Al-Qurashi, A., N. McIntyre, H. Wheater, and C. Unkrich, 2008, Application of the Kineros2 rainfall-runoff model to an arid catchment in Oman: Journal of Hydrology, v. 355, p. 91-105. Al-Rawas, G. A., and C. Valeo, 2008, Issues with Flash Flood Modeling in the Capital region of Sultanate Oman, Geomatics Engineering, Schulich School of Engineering, University of Calgary, p. 1-11. 37

Arnold, J. G., R. Srinivasan, R. S. Muttiah, and J. R. Williams, 1998, Large area hydrologic modeling and assessment part I: Model Development 1: JAWRA Journal of the American Water Resources Association, v. 34, p. 73-89. Beven, K., 1984, Infiltration into a class of vertically non-uniform soils: Hydrological Sciences Journal-Journal Des Sciences Hydrologiques, v. 29, p. 425-434. Beven, K., 1995, Linking Parameters across Scales - Subgrid Parameterizations and Scale-Dependent Hydrological Models: Hydrological Processes, v. 9, p. 507525. Beven, K., 2001, Rainfall-runoff modelling: the primer: Chichester, John Wiley & Sons, Ltd, 355 p. Boughton, W., and F. Chiew, 2007, Estimating runoff in ungauged catchments from rainfall, PET and the AWBM model: Environmental Modelling & Software, v. 22, p. 476-487. Bronstert, A., 2004, Rainfall-runoff modelling for assessing impacts of climate and land-use change: Hydrological Processes, v. 18, p. 567-570. Chow, V. T., D. R. Maidment, and L. W. Mays, 2005, Applied hydrology, McGrawHill New York. Croke, B. F. W., F. Andrews, A. J. Jakeman, S. M. Cuddy, and A. Luddy, 2006, IHACRES classic plus: a redesign of the IHACRES rainfall-runoff model: Environmental Modelling & Software, v. 21, p. 426-427. DHI, 1993, Mike11, A microcomputer-based modelling system for rivers and channels, User and reference manual, in D. H. Institute, ed., Hrsholm, Denmark.

38

Dye, P. J., and B. F. W. Croke, 2003, Evaluation of streamflow predictions by the IHACRES rainfall-runoff model in two South African catchments: Environmental Modelling & Software, v. 18, p. 705-712. Freiwan, M., and M. Kadioglu, 2008, Climate variability in Jordan: International Journal of Climatology, v. 28, p. 69-89. Gheith, H., and M. Sultan, 2002, Construction of a hydrologic model for estimating Wadi runoff and groundwater recharge in the Eastern Desert, Egypt: Journal of Hydrology, v. 263, p. 36-55. Hadadin, N., M. Qaqish, E. Akawwi, and A. Bdour, 2010, Water shortage in Jordan Sustainable solutions: Desalination, v. 250, p. 197-202 Huffman, G. J., R. F. Adler, D. T. Bolvin, G. J. Gu, E. J. Nelkin, K. P. Bowman, Y. Hong, E. F. Stocker, and D. B. Wolff, 2007, The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales: Journal of Hydrometeorology, v. 8, p. 38-55. Huffman, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001, Global precipitation at one-degree daily resolution from multisatellite observations: Journal of Hydrometeorology, v. 2, p. 36-50. Hughes, D. A., R. Institute for Water, and C. South Africa. Water Research, 1997, Southern African" FRIEND": The Application of Rainfall-runoff Models in the SADC Region, Water Research Commission Pretoria, South Africa. Jakeman, A. J., and G. M. Hornberger, 1993, How much complexity is warranted in a rainfall-runoff model: Water Resources Research, v. 29, p. 2637-2649.

39

Kilpatrick FA, Cobb ED, 1985, Measurement of Discharge Using Tracers, Chapter A16, Techniques of Water-Resources Investigations, Book 3, Vol 52: U.S. Geological Survey T. Kotwicki, V., 1987, International-symposium on water for the future - hydrology and water-resources development in perspective: Hydrological Sciences JournalJournal Des Sciences Hydrologiques, v. 32, p. 420-421. Lange, J., C. Liebundgut, and A. P. Schick, 2000, The importance of single events in arid zone rainfall-runoff modelling: Physics and Chemistry of the Earth Part B-Hydrology Oceans and Atmosphere, v. 25, p. 673-677. Lin X, 1999, Flash floods in arid and semi-arid zones. In Hydrology TDi (ed): UNESCO, Paris. Matson, K. C., C. Stallings, Jennings, M. G.D., R.A., and S. W. Coffey, 1995, Watershed scale hydrological modelling in a GIS environment: GIS/LIS '95 Conference Exposition, p. 14-16. McIntyre, N., and A. Al-Qurashi, 2009, Performance of ten rainfall-runoff models applied to an arid catchment in Oman: Environmental Modelling & Software, v. 24, p. 726-738. Osborn, H. B., K. G. Renard, and J. R. Simanton, 1979, Dense Networks to Measure Convective Rainfall in the Southwestern United-States: Water Resources Research, v. 15, p. 1701-1711. Perrin, C., C. Michel, and V. Andrassian, 2003, Improvement of a parsimonious model for streamflow simulation: Journal of Hydrology, v. 279, p. 275-289. Pilgrim, D. H., T. G. Chapman, and D. G. Doran, 1988, Problems of Rainfall-Runoff Modeling in Arid and Semiarid Regions: Hydrological Sciences JournalJournal Des Sciences Hydrologiques, v. 33, p. 379-400.

40

Reed, S., J. Schaake, and Z. Y. Zhang, 2007, A distributed hydrologic model and threshold frequency-based method for flash flood forecasting at ungauged locations: Journal of Hydrology, v. 337, p. 402-420. Rodier J, Roche M, 1978, River flow in arid regions. Hydrometry: Principles and Practices:453. Safriel, U. N. 2006 Dryland development, desertification and security in the Mediterranean. In Desertification in the Mediterranean Region. A Security Issue (ed. Kepner, W. G., Rubio, J. L., Mouat, D. A. & Pedrazzini, F.), Springer, 3, 227250. Scharffenberg, W., and M. Fleming, 2010, Hydrologic modeling system HEC-HMS v3.2 users manual.: Davis, USA, USACE-HEC. Schoener, G., 2010, Comparison of AHYMO and HEC-HMS for Runoff Modeling in New Mexico Urban Watersheds, University of New Mexico, New Mexico. Shadeed, S., and M. Almasri, 2010, Application of GIS-based SCS-CN method in West Bank catchments, Palestine: Water Science and Engineering, v. 3, p. 113. Shah, S. M. S., P. E. O'Connell, and J. R. M. Hosking, 1996, Modelling the effects of spatial variability in rainfall on catchment response. 1. Formulation and calibration of a stochastic rainfall field model: Journal of hydrology, v. 175, p. 67-88. Sharma, K. D., and J. S. R. Murthy, 1996, Ephemeral flow modelling in arid regions: Journal of Arid Environments, v. 33, p. 161-178. Sherif, M. M., M. M. Mohamed, A. Shetty, and M. Almulla, 2011, Rainfall Runoff Modeling of Three Wadis in the Northern Area of UAE: Journal of Hydrologic Engineering, v. 16, p. 11.

41

Sorooshian, S., K. L. Hsu, X. Gao, H. V. Gupta, B. Imam, and D. Braithwaite, 2000, Evaluation of PERSIANN system satellite-based estimates of tropical rainfall: Bulletin of the American Meteorological Society, v. 81, p. 2035-2046. Ushio, T., K. Sasashige, T. Kubota, S. Shige, K. Okamoto, K. Aonashi, T. Inoue, N. Takahashi, T. Iguchi, M. Kachi, R. Oki, T. Morimoto, and Z. I. Kawasaki, 2009, A Kalman Filter Approach to the Global Satellite Mapping of Precipitation (GSMaP) from Combined Passive Microwave and Infrared Radiometric Data: Journal of the Meteorological Society of Japan, v. 87, p. 137-151. Wheater, H. S., A. P. Butler, E. J. Stewart, and G. S. Hamilton, 1991, A Multivariate Spatial-Temporal Model of Rainfall in Southwest Saudi-Arabia.1. Spatial Rainfall Characteristics and Model Formulation: Journal of Hydrology, v. 125, p. 175-199. Wheater HS, Jakeman AJ, Beven KJ, 1993, Progress and directions in rainfall-runoff modelling. In A.J. Jakeman MBBaMJM (ed), Modelling Change in Environmental Systems. Chichester: Wiley, pp 101-132. Xie, P. P., J. E. Janowiak, P. A. Arkin, R. Adler, A. Gruber, R. Ferraro, G. J. Huffman, and S. Curtis, 2003, GPCP Pentad precipitation analyses: An experimental dataset based on gauge observations and satellite estimates: Journal of Climate, v. 16, p. 2197-2214.

42

Appendix A
Abushandi E. & Merkel B. 2010 Application of IHACRES rainfallrunoff model to the Wadi Dhuliel arid catchment, Jordan. J. Water and Climate Change (Published)
Eyad H. Abushandi (corresponding author) Broder J. Merkel Institute for Hydrogeology, TU Bergakademie Freiberg, Gustav-Zeuner-Str. 12, 09596 Freiberg/Sa, Germany Tel.: +49 3731 39 3118 Fax: +49 3731 39 2720 E-mail: [email protected]

Abstract
With increasing stress on water resources in Jordan, application of rainfall-runoff models can be part of the solution to manage and sustain the water sector. In this paper, the metric conceptual IHACRES model is applied to the Wadi Dhuliel arid catchment, north-east Jordan. Rainfall-runoff data from 19 storm events during 1986 to 1992 have been used in this study. Flood estimation was performed on the basis of daily scales and storm events scales. The model was extended for snowfall in order to cope with such extreme events. Although the best performance of the IHACRES model on a daily basis is poor, the performance on storm events scale showed a good agreement between observed and simulated streamflow. Apart from model parameter values, the principal reasons for IHACRES model success in this region are thought to be based on antecedent soil moisture conditions, rainfall duration and rainfall intensity before and during each storm. The model outputs were likely to be sensitive when the 43

monitored flood was relatively small. The optimum parameter values were influenced by the length of calibration data and event specific changes.
Key words: arid regions, effective rainfall, IHACRES, rainfall-runoff, snow

Introduction

Around 95% of Jordan lies on arid and semi-arid terrains between the latitudes of 2911 N and 3322 N and the longitudes of 3459 E and 3912 E. Water scarcity is the major problem in Jordan, which has one of the worlds fastest growing populations with a growth rate of 4.5% annually (Al-Halasah and Ammary 2006; Jaber et al 1997). Due to both population increase and rising of general prosperity the use of water resources in Jordan is dramatically increasing. Sporadic and unpredictable severe rainstorm events, flash floods and droughts are the main characteristics of the hydrological system in Jordan. Only 9% of Jordans territory receives more than 200 mm of the annual rainfall (Al-Halasah and Ammary 2006). In comparison to some of Jordans neighbouring countries, Jordan has the lowest annual water share. Therefore, an urgent need to focus on water resource sustainability management is essential. With the exception of the Yarmouk River, surface water flow in Jordan is extremely rare and occurs in the following three ways: 1 a few creeks fed from groundwater through springs; 2 streamflow or floodwater formed after significant rainstorms in winter; 3 discharge of treated wastewater. The average annual floodwater for the entire Jordan area is 255 106 m3 (AlHalasah and Ammary 2006), which contributes 3% of the total annual rainfall. Although the quantity of floodwater is relatively small, it might increase the water availability at the catchment scale. There are a few short-term projects on flood flow 44

measurements but long-term flow data are generally not available for arid sites in Jordan. The major reason for this situation is a lack of awareness of the importance of monitoring flood flows in the governmental authorities. In addition, even in the rare cases where gauging stations exist, general measurement problems occur during intense flood events reducing the quality and completeness of the data (Lange et al 2000). Thus for arid regions in general and for Jordan in particular two requirements can be stated: 1 improved monitoring systems for flood flows; 2 evaluation tools based on limited data. Rainfall-runoff models have been widely used for the last 50 years (Nash 1959) as a means of describing hydrological processes. Hundreds of rainfall-runoff models have been developed throughout the world, especially in Europe to provide river flow forecasting (Beven 2001). However, most of these models are designed to estimate runoff from persistent rainfall (Pilgrim et al 1988), making most of them appropriate tools for humid environments only. Arid and semi-arid areas have particular challenges that have received less attention (Wheater et al 2008). A comprehensive review of modelling rainfall-runoff in arid and semi-arid regions and related difficulties was made by Pilgrim et al. (1988). Because of harsh climatological and physical conditions, modelling the rainfall-runoff relationship has been the dilemma for hydrologists working in arid regions. Obviously, only a limited number of rainfall-runoff models have been developed and modified for arid and semi-arid environments. Perhaps the first published research on characterizing the rainfall-runoff relationship in a semi-arid area was given by Osborn & Hickok in 1968. They analysed the effect of rainfall variability in producing runoff for Alamogordo Creek watershed in New Mexico.

45

Lane & Renard (1971) evaluated stochastic modelling for ephemeral runoff from Walnut Gulch semi-arid experimental watershed in Arizona. The model was based on five statistical distributions of actual flow events to generate synthetic individual flows. The basic assumption of linear correlation between the starting date of the wet season and the number of thunderstorm runoff events per season was considered. The authors concluded that the length of runoff season plays an important role in the variability of the results. Ultimately, the typical application of rainfall-runoff models to arid and semi-arid regions is limited to parametric models (Drissel & Osborn 1968; Osborn & Hickok 1968; Lane & Renard 1971; Croke & Jakeman 2008; McIntyre & Al-Qurashi 2009). Although it is recognized by many hydrologists that the rainfall-runoff models cannot give a full picture or plethora of reality (Beven 1984; Pilgrim et al. 1988; Bronstert 2004; Al-Qurashi et al. 2008; McIntyre & Al-Qurashi 2009), modelling the rainfall-runoff relationship is a very useful tool to predict streamflow in arid catchments. In general, the choice of rainfall-runoff model is based on three main roles: 2 The availability and the quality of hydrological data, especially rainfall and runoff data; 3 Model structure and model ability of regionalization; 4 Catchment characteristics and flow homogeneity. These roles can be slightly different depending primarily on the catchment type and secondarily on the catchment size. For instance, the challenge of modelling urbanizing catchments is how to link the population growth and the effect of growing in industrialization. Aucharova & Khomich (2006) have mentioned that the type and concentration of urban runoff contaminants are primarily based on seasonal period

46

and type of functional zones, therefore it is necessary to consider this conclusion in modelling urbanizing catchments. Attention has been paid to the effect of urbanization proceeds and manmade drainage system characteristics on the variability of runoff magnitudes (Rao et al. 1971; Hall 1977). In addition urban rainfall-runoff models require more spatial details of overland flow, rain-dependent inflow and infiltration (Dongquan et al. 2009). On the other hand, applications of rainfall-runoff models for agricultural catchment have to deal with specific problems such as land degradation (Bulygina et al 2009), leaching fertilizers and their impact on surface and ground water quality (Sten Bergstrm & Brandt 1987) and soil erosion assessment (Lopezbermudez 1990; Hrissanthou et al. 2003). This usually requires some implementations or modifications of rainfall-runoff models. The IHACRES rainfall-runoff model has been successfully applied to many arid and semi-arid catchments all over the world (e.g. Australia, Oman, Jordan and many parts of Africa). IHACRES stands for Identification of unit Hydrograph And Component flows from Rainfall, Evaporation and Streamflow. It has been used to predict streamflow in ungauged catchments (Kokkonen et al 2003), study land cover effects on hydrologic processes (Croke and Jakeman 2004; Kokkonen et al 2001) and investigate dynamic response characteristics and physical catchment patterns (Kokkonen et al 2003; Sefton and Howarth 1998). In a particular study IHACRES has been applied to the Wadi Faynan area of south-western Jordan as part of the Water Life and Civilization Project. Monthly flow data from the River Jordan gauging station in conjunction with monthly rainfall data from Mt Kenaan, Israel, were used to apply IHACRES for the period 19881993. A

47

calibrated version of IHACRES was able to explain 96% of flow records variability (Whitehead et al 2006). The Zarqa River is the second-largest river in Jordan after the Yarmouk River. The total area of the river drainage network is around 3882 km2 (the total area may differ slightly from one author to another). It is divided into two sub-basins: the Seil Al-Zarqa sub-basin and the Wadi Dhuliel sub-basin with a lower outlet near the King Talal Dam (Figure.A 1). The river receives water from two main sources: rainwater during winter, contributing approximately 47.4% of the annual yield, and approximately 52.6% due to treated wastewater from the Khirbet Al-Samra, AlBaqaa, Jarash, Abu-Nsair and Almafraq Wastewater Treatment Plants. In addition, there is a very limited contribution of groundwater through surrounding springs. Unfortunately this amount of groundwater contribution is not well cited.

Figure.A 1| Location map of Al-Zarqa basin: including Seil Al-Zarqa

sub-basin and Wadi Dhuliel sub-basin (map datum: WGS 84).

48

Wadi Dhuliel catchment is an appropriate example of an arid catchment and an important source for passing floodwater from the northern part into the Al-Zarqa River. In comparison to other hydrological models, the IHACRES model only requires three data sets (rainfall, temperature and streamflow) per time unit and a small number of parameters. This, however, may overcome the problem of the observed data scarcity for arid regions. In addition the structural simplicity of the IHACRES model assists the good performance of the model for many types of catchment. This paper examines the application of the IHACRES model into 19 rainstorm events that caused streamflow obtained from the Wadi Dhuliel sub-basin by using daily rainfall, temperature and streamflow data to evaluate the accuracy of the calibration method in that region.

2
2.1

Methodology
IHACRES model structure

Model developers divide rainfall-runoff models into three categories: metric, conceptual and physics-based models (Beck 1991; Croke et al 2006; Kokkonen et al 2001). Metric rainfall-runoff models are the simplest models based on observed data including rainfall and runoff records to characterize the catchment interaction. Conceptual models describe many internal aspects to characterize the catchment interaction. Physics-based models couple mathematical-physical theories and flow equations (Navier-Stokes) to achieve precise simulations. IHACRES (Jakeman and Hornberger 1993; Jakeman et al 1990) is a hybrid conceptual-metric model, using the simplicity of the metric model to reduce the parameter uncertainty inherent in hydrological models (Croke and Jakeman 2005).

49

The main objective of IHACRES is to characterize catchment-scale hydrological behaviour using as few parameters (often about six parameters) as possible (Littlewood 2003). Therefore, it can successfully be applied to arid regions where hydrological data sets are very rare. The original structure of IHACRES uses the exponential soil moisture drying rate index. Several versions of IHACRES have recently been developed and improved to achieve a good simulation of ephemeral streams in arid regions. The model contains a non-linear loss module followed by a linear module; the non-linear loss module converts rainfall into effective rainfall, while the linear module transfers effective rainfall to streamflow. Figure.A 2 shows the model components.

Figure.A 2| Generic structure of IHACRES model.

In this study the classic redesigned IHACRES version (Croke et al 2006; Jakeman and Hornberger 1993) has been used. The non-linear loss module converts rainfall (rk ) into an effective rainfall (u k ) by considering both the infiltration rate and evapotranspiration. In order to obtain the effective rainfall, a catchment wetness index or antecedent precipitation index, representing catchment saturation, is calculated for

50

each time step. Usually, a non-linear loss module within IHACRES includes three parameters: c is the adjustment parameter, f is a temperature modulation parameter and w(const ) is the rate at which catchment wetness declines in the absence of rainfall. The initial stage is to determine the drying rate w , and the catchment moisture index S k at each time step, which is given by:
w = w( const ) e ( 20 t k ) f

(1)

where w is the drying rate at each time step, w (const ) is the rate at which catchment wetness declines in the absence of rainfall, tk is the temperature at time step k and f is a temperature modulation parameter (C1), which determines how w changes with temperature. Catchment wetness index S k is computed for each time step on the basis of recent rainfall and temperature records. The loss module is used to account for the effect of antecedent weather conditions on the current status ( S k ) of soil moisture and vegetation conditions, and evapotranspiration effects (Schreider et al 1995).
S k = c rk + (1 1 ) S k 1 w( k )

(2)

where c is the adjustment parameter and controls the amount by which S k is increasing by a rainfall event (Post and Jakeman 1999), rk is the rainfall at time step k. Finally the effective rainfall (uk ) in the model is given by: u k = rk s k if rk s k 0 . Since the study area is characterized by ephemeral streams where there is no runoff if there is no rainfall, computing effective rainfall if rk is >0 for ephemeral streams can make equation (3) true, both physically and dimensionally. In the linear (3)

51

routing module, the effective rainfall is converted into streamflow (Qk ) . The storage configurations of two parallel storage components have been applied. The linear model employs discrete-time intervals, transfer function and a representation of the Unit Hydrograph (UH).
Q(qk ) = qQ(qk 1) + qu( k )
Q(sk ) = sQ(sk 1) + su( k )

(4) (5)

where (Q(qk ) , Q(sk ) ) are quick and slow streamflow components. Delta in the u ( k ) is the delay between rainfall and streamflow response. The parameters ( q , s ) are the recession rates for quick and slow storage, whereas the parameters ( q , s ) represent the fraction of effective rainfall. The UH of total streamflow is the total of both quick and slow flow UHs. As an external parameter, snow melt parameter (Cmelt) was added into the q fraction in the case of snowfall events:

Q(qk) = qQ(qk 1) + (q )(Cmelt )u(k )

(6)

The snow melt parameter (Cmelt) will be automatically blocked if there are no snowfall events; it is actually connected to the temperature gradient. In other words; the Cmelt parameter is not functioning if the temperature is higher than the freezing degree. To predict flood flow for ungauged catchments by means of calibrated rainfall-runoff models, a technique for estimating parameter values is required. There are numerous studies for determining these values. (Merz and Bloschl 2004) estimated parameter values based on regression analysis between proposed parameter values and catchment attributes. However, this regression application requires parsimonious models that have strong relationships between parameter and catchment attributes (Croke and Littlewood 2005). In addition, the Nash-Sutcliffe efficiency (Ef) 52

has frequently been applied to assess the goodness of modelled streamflow to observed records:
Ef

(q q ) =1 (q q )
i i i

(7)

i is modelled streamflow and q is the mean value where qi is observed streamflow, q

of observed streamflow. The Ef will be used in this study to estimate the goodness of model outputs. In order to measure the reliability of simulated values the Relative Error of Estimate was calculated.:
1 M

(Q
1

sim

Qobs ) 2 100

REE =

Q obs

(8)

2.2

Study area and available data

Wadi Dhuliel has always played a key role in increasing water levels in the Al-Zarqa River and being a valuable source of natural groundwater recharge. It is located in the north-east part of Jordan. Most of the catchment area belongs to the Al-Zarqa Basin. Around 10% of the upper part of Wadi Dhuliel catchment is situated on Syrian territory (Figure.A 3). The area is characterized by a very gently undulating topography with an elevation varying from 512 m in the south-west to 1400 m in the north. The area of interest is approximately 1985 km2.

53

Figure.A 3| Wadi Dhuliel sub-basin: location map, drainage network and rainfall

and discharge stations. DTM based on 90 m SRTM.

Besides Wadi Dhuliel, there are a number of wadis in the catchment area such as Wadi Al-Zaatari, Wadi Al-Ajeb and Wadi Alghar. The flow direction is from the north towards Seil Al-Zarqa sub-basin in the south-west. Hydrological features of the Wadi Dhuliel catchment are disturbed by a strong influence of agricultural activities and high pressure on groundwater resources. The area of interest is characterized by warm and dry summers (May to September) and moderate cold and wet winters (October to April). The overall annual rainfall is 123 mm on average (Figure.A 4). The average annual temperature is 16.8 C (19762005). Table.A1 shows average temperature and standard deviation

54

during the driest and wettest months. The data for this table were compiled by Almafraq meteorological station at an elevation of 675 m above sea level.

Figure.A 4| Climograph for Almafraq Meteorological Station/north-east

Jordan, the period from 19762005 (Jordan Meteorological Department).

Table.A 1| Temperature description for the period of 19762005 Jan. Feb. Mar. Average (C) Average maximum (C) Average minimum (C) PET (mm)* Apr. May Jun.

7.4 12.7 2.1 145

Jul.

8.6 14.2 2.9 185

Aug.

11.5 18 5.1 310

Sep.

16.2 24 8.5 566

Oct.

20.4 28.7 12.0 842

Nov.

22.9 31.6 14.1 1034

Dec.

Average (C) Average maximum (C) Average minimum (C) PET (mm)*

24.7 33.1 16.32 1186

24.6 33 16.31 1181

23.1 31.3 14.9 1054

19.5 27.2 11.73 779

13.6 20.2 6.92 415

9.0 14.5 3.51 204

* Potential evapotranspiration (PET) is calculated by the Thornthwaite equation.

55

The annual rainfall magnitudes show a distinct west-east gradient from the relatively wet west regions with about 600 mm to the Jordanian desert or Al-Badia with less than 100 mm. Based on average annual precipitation (P) and potential evapotranspiration (PET) for the period of 1976 to 2005 of Almafraq Meteorological Station (Figure.A 5), the degree of aridity was calculated by the UNEP Aridity Index (Al) (Safriel 2006).

Figure.A 5| UNEP Aridity Index for the study area. Dashed line represents

the aridity index of Almafraq Meteorological Station.

IHACRES model requires three data sets per time unit: (a) rainfall, (b) streamflow for calibration purpose and (c) temperature or potential

evapotranspiration.

56

Typically, data sets for arid areas are limited to daily rainfall, temperature and in lucky cases short periods of flood flow records. The data sets used in this research were derived from the Surface Water Resources Unit at the Jordan Ministry of Water and Irrigation (JMWI) and Jordan Meteorological Department (JMD). The initial step was a careful screening with respect to errors and missing values in the datasets. Daily rainfall data are available from seven rainfall gauging stations, which are fairly well distributed over the entire catchment area. Recently, new rainfall gauges were added; however, their recording period is less than five years. Almafraq station has the highest annual rainfall (158 mm) while Qasr Al-Hallabat station has the lowest rainfall (79.2 mm). Seventy-three per cent of the annual rainfall occurs during November, December, January and February. An average rainfall was calculated from seven rainfall and meteorological stations. Table.A 2 presents detailed rainfall information and period of records for each station.

57

Table.A 2| Detailed information about rainfall stations utilized for the study
Station Station code on the code in map (Fig. JMWI 3) (mm) name in metres) precipitation records (years) Station Height (asl annual Period of period Mean Total of

Sabha and A B C AL0058 Subhiyeh AL0059 AL0048 Um-Jimal * Al-Khaldiya Wadi D AL0055 Dhuliel Nursery G F E
* a b

843 670 600

108.1 119.3 125.4

19682002 19692002 19682002

35 35 35

580

137

19682002

35 a

No Code AL0049

Almafraq *b Qasr Al-

675 590

158 79.2 135.3

19752005 19682002 19682002

30 35 52

hallabat AL0012 Sukhnah 556

Meteorological station. Missing data in the years 19681971. Almafraq station is the only meteorological station from Jordan Meteorological Department, all others belong to

the Jordan Ministry of Water and Irrigation.

Since availability of daily temperature records was limited, daily temperature data were taken from Um-Aljimal Meteorological Station only. The station is located at the upper part of Wadi Dhuliel catchment with an elevation of 676 m asl. Daily stream flow records were collected between 1986 and 1992 from Wadi Al-Zaatari streamflow gauging station close to the junction with Wadi Al-Dhuliel. This flow record has been used to test IHACRES performance during the calibration stage. Table.A 3 provides a very brief summary of the flow measurements.

58

Table.A 3| Summary of discharge from field measurements (JMWI)

Cross Mean depth section (m) width (m) (m) (m2) (m s1) Max depth section area velocity reading (m) (m3 s1) Cross Mean Gauge discharge Observed

6.00 36.00 35.80 38.00 35.80

0.16 0.18 0.90 0.91 1.20

0.35 0.27 1.00 1.05 1.36

0.98 6.38 32.30 34.6 43.00

0.478 1.342 0.786 1.066 1.102

0.02 0.225 0.845 0.91 1.15

0.469 8.56 25.4 36.9 47.4

The

stage-discharge

relation

was

developed

from

direct

discharge

measurements carried out by the JMWI team by means of a current meter during the first and the second significant rainstorms in November 1986 and January 1987. Figure.A 6 shows the developed rating curve from a series of observed flow measurements plotted versus gauge heights at each time step. Most often the discharge in this area is characterized by flash flood waves of very short durations ranging from a few hours to five days within an average of three days. The discharge varies from Q = 0 to Q = 125.2 m3 s1.

59

Figure.A 6| The rating curve for Wadi Al-Zaatari in the period of 1986

1987.

Application of IHACRES was limited to the flow measurements period. As explained by (Hansen et al 1996) effective rainfall ( u k ) is calculated from the accumulated rainfall depth ( rk ) in the non-linear loss module at each time step, where the catchment wetness index S k values at this stage represent soil moisture characteristics in arid regions and runoff generation thresholds. To understand soil moisture characteristics, additional data sets were collected from a recently installed weather station in the study area (since January 2008). The data included rainfall, temperature and soil moisture. The point was to study the changes of rainfall and temperature in parallel with soil moisture changes. Since soil physical properties are fixed (i.e. infiltration rate), the potential values of the catchment wetness index S k for the years between 1986 and 1992 were adjusted at both daily and storm event scales.

60

Simple linear interpolation method was used to develop daily and storm event time series of soil moisture for the years between 1986 and 1992 based on the additional data sets.

Results
Flood flow records from the Al-Zaatari gauging station show the main

characteristic of surface water flow in the area. The runoff coefficient in the study area was 2.3% on average. Some flood flow events show more than one flow peak during a single rainstorm (Figure.A 7) and are characterized by steep rising peaks. The results show that the catchment tends to have only very few flood flow events (19 events during the period of 06.11.1986 to 11.02.1992, with an average of 3 storms per year).

Figure.A 7| Representative behaviours of continuous streamflow: one day (29

November 1986) and five days streamflow (08 November14 November 1986) respectively.

The IHACRES model was calibrated by using the Al-Zaatari daily flood flow records over a period from 06.11.1986 to 11.02.1992. The best parameter values

61

containing the best fit between modelled and observed streamflow are listed in Table.A 4. Table.A 4| Optimized IHACRES parameter values
C 0.002

w( const)
120

F 0.09

q
0.07

q
0.04

s
1

s
0

Cmelt 0.1

In general, it is unusual for north-east parts of Jordan to receive a heavy snowfall. However, the snowstorm that began in January 1990 and ended in February of the same year was heavy enough to reach the arid parts crossing the Wadi Dhuliel arid catchment. This snowstorm changed the streamflow magnitudes and behaviour in 1990. A fraction of snow melt parameter (Cmelt) was added for such particular events to reduce the quick fraction ( q ) influence. The total flood magnitude from each rainstorm was calculated as a single unit by the mean of total streamflow of each day recorded. The application of IHACRES to both daily and storm events basis gave a better view of streamflow behaviour in the catchment. The goodness of fit between observed flood flow ( qi ) and modelled flood i ) was estimated using the Nash-Sutcliffe efficiency ( E f ). flow ( q A good performance of IHACRES was obtained with E f equal to 0.86 on the storm events scale, while E f was 0.43 on the daily scale using the same parameter values. This, however, shows a poor performance of the IHACRES model on daily basis application. The result of IHACRES simulations in comparison to monitored data on both daily and storm events scales are shown in Figures.A 8 and 9.

62

Figure.A 8| Observed and simulated flood flow on a daily basis.

63

Figure.A 9| Observed and simulated flood flow based on storm events.

The average observed flood on a daily basis was 0.14 mm day1 while the average of simulated flood was 0.16 mm day1. In contrast, the average observed flood per storm event was 0.65 mm storm1 while the average of simulated floods of all storms was 0.62 mm storm1. As expected from E f the Relative Error of Estimate of daily simulated-observed streamflow was 1.45% while in storm events the basis was 0.44%.

Discussion
Rainfall behaviour tends to be asymmetric in both space and time, thus

affecting the streamflow magnitudes. For this reason, arid catchments are more amenable to simplified models. Because IHACRES is a parametric efficient rainfallrunoff model it is applicable in arid areas, which are dominated by rapid responses to weather variables. Since the model is a lumped model, it has the capability to avoid 64

the spatial variability of the rainfall and streamflow. Furthermore the model requires only a few input data. In general, rainfall-runoff models have many limitations and aim to achieve a high accuracy. Errors during simulations often occur because of missing data, or complexity of hydrological behaviour. The time step calculation is very important for IHACRES modelling in an arid region. As the results show, it is crucial to change the calculations from daily time steps to storm event time steps at the same parameter values due to the dependency of the present value on previous records. Apart from IHACRES structure there are many other reasons that may increase the relative error such as the flood fluctuation from one storm to another or stage-discharge errors. The Relative Standard Error is likely to be significant when the monitored flood is relatively small. During the rainy season of the year 1990, the residuals were extremely high in both daily and storm event scales, which can be explained by two reasons: the occurrence of snow instead of rain, and the relatively very small monitored streamflow (0.04 mm storm1). This snowstorm changed the entire behaviour of rainfall and temperature gradients. In this particular case the IHACRES model converted the total precipitation (including snow) into effective rainfall and then into streamflow through a quick flow linear module. The identification of optimum model parameters at this specific event was difficult; therefore, a snow melt parameter was added to identify snow meltrunoff conditions. This extreme event happened only once during the last 50 years as it is recorded, therefore it was challenging to study the IHACRES model simulation abilities under such conditions. Relatively, the period of observed and simulated data is not short as in many other related studies. Therefore, it can provide a higher

65

accuracy of the calibration parameters which might be extended to test streamflow for the periods where no observed data are available. Particular attention has been paid to the rainfall patterns after the snow storm in 1990 (Figure.A 10). Tracking the rainfall rate shows that the average annual rainfall across the entire Wadi Dhuliel catchment is 112.7 mm for the years 19902008 while the average annual rate for the years 1969 2002 is 123.2 mm. This generally shows that the annual rainfall totals are declining but there have been no snowfall events so far.

Figure.A 10| Annual rainfall rate measured in Wadi Dhuliel Catchment for the

years 19902008. Dashed line represents the mean annual rainfall of the same period.

66

Based on the land use map derived from a Landsat Thematic Mapper (TM) imagery of the year 1987 (Figure.A 11) the agricultural area contributes approximately 36.2% from the total catchment area. Due to this fact, significant quantities of agrochemicals are expected to be transferred from the Wadi Dhuliel catchment to the Seil Al-Zarqa catchment, and end in the King Talal Dam. Therefore it is necessary to predict the potential contaminations in future research.

Figure.A 11| Land use map of Wadi Dhuliel sub-basin, Landsat Thematic

Mapper (TM) of the year 1987, 30 m resolution.

Moreover, application of rainfall-runoff models in Jordan is a challenging task because of data scarcity and the complex interaction between climate variables. Apart from applying IHACRES rainfall-runoff model for the first time to arid parts of Jordan, the paper added three new aspects:

67

1 Adding and testing a new parameter in IHACRES model, the snow melt parameter, because arid regions do not have snowfall events. 2 Dividing the data into daily and storm event simulation, not just daily simulation. 3 Taking advantage of a data set obtained recently to correlate historical data, i.e. the soil moisture data set from our weather station.

Conclusion

The IHACRES rainfall-runoff model is applicable in the Jordanian arid areas, which are dominated by ephemeral streams and rapid responses to storm events. According to the obtained results, the model is able to adequately simulate streamflow in arid catchments when applying the model on a storm event scale. Dividing the same data into daily and storm event was a good option to evaluate IHACRES at different scales. A good performance of the IHACRES model showed E f equal to 0.86 on the storm events scale, while E f was 0.43 on a daily scale. The results therefore depend on the chosen time interval. A new parameter (snow melt) has been added into the IHACRES model. By adding a snow melt parameter the model could be enforced significantly. Changes in rainfall and temperature significantly affect soil moisture capacity, therefore calibration methods of IHACRES models need to be extended. The extension of calibration methods might include automated techniques instead of manual. Longer calibration periods are needed in order to reduce the uncertainty in model parameters and predict both climate change and anthropogenic impact on the study area. A key role of runoff generation can be concluded in four points: (1) antecedent soil moisture conditions, (2) rainfall intensity, (3) rainfall amounts before and during streamflow events and (4) the period of rainfall storms which affect soil storage capacity.

68

Acknowledgment
The authors are thankful to the Jordan Ministry of Water and Irrigation, particularly the Surface Water Unit, and the Jordan Meteorological Department for providing the necessary data sets.

References
Al-Halasah, N. K. & Ammary, B. Y. 2006 Water resources policy and management in Jordan. In Integrated Urban Water Resources Management. Proceedings of the NATO Advance Research Workshop, 1923 October 2005, pp. 5968, Senic, Slovak Republic. Al-Qurashi, A., McIntyre, N., Wheater, H. & Unkrich, C. 2008 Application of the Kineros2 rainfall-runoff model to an arid catchment in Oman. J. Hydrol.
355(14), 91105.

Aucharova, A. & Khomich, V. 2006 Urban runoff contamination, problems of treatment and impact on receiving water. In Integrated Urban Water
Resources Management. Proceedings of the NATO Advance Research

Workshop, 1923 October 2005, pp. 191200, Senic, Slovak Republic. Beck, M. B. 1991 Forecasting environmental change. J. Forecas. 10(12), 319. Bergstrm, S. & Brandt, M. 1987 Simulation of runoff and nitrogen leaching from two fields in southern Sweden. Hydrol. Sci. J. 32(2), 191205. Beven, K. 1984 Infiltration into a class of vertically non-uniform soils. Hydrol. Sci. J.
29(4), 425434.

Beven, K. 2001 Rainfall-runoff Modeling The Primer. John Wiley and Sons, Ltd, Chichester, UK, pp. 169.

69

Bronstert, A. 2004 Rainfall-runoff modeling for assessing impacts of climate and land-use change. Hydrol. Process. 18(3), 567570. Bulygina, N., McIntyre, N. & Wheater, H. 2009 Conditioning rainfall-runoff model parameters for ungauged catchments and land management impacts analysis.
Hydrol. Earth Sys. Sci. 13(6), 893904.

Croke, B. F. W. & Jakeman, A. 2005 A catchment moisture deficit module for the IHACRES rainfall-runoff model. Environ. Model. Soft. 20(7), 977977. Croke, B. F. W. & Jakeman, A. J. 2008 Use of the IHACRES rainfall-runoff model in arid and semi-arid regions. In Hydrological Modeling in Arid and Semi-arid
Areas (ed. Wheater, H. S. Sorooshian & Sharma, K. D.), Cambridge

University Press, Cambridge, pp. 4148. Croke, B. F. W. & Littlewood, I. G. 2005 Comparison of alternative loss modules in the IHACRES model: an application to 7 catchments in Wales. In
International Congress on Modeling and Simulation (ed. Zerger, A. & Argent,

R. M.), MODSIM2005, Modeling and Simulation Society of Australia and New Zealand, Melbourne, Australia, pp. 29042910. Croke, B. F. W., Andrews, F., Jakeman, A. J., Cuddy, S. M. & Luddy, A. 2006 IHACRES classic plus: a redesign of the IHACRES rainfall-runoff model.
Environ. Model. Soft. 21(3), 426427.

Dongquan, Z., Jining, C., Haozheng, W., Qingyuan, T., Shangbing C., & Zheng, S. 2009 GIS-based urban rainfall-runoff modeling using an automatic catchmentdiscretization approach: a case study in Macau. Environ Earth Sci. 59, 465 472.

70

Drissel, J. C. & Osborn, H. 1968 Variability in rainfall producing runoff from a semiarid rangeland watershed, Alamogordo Creek, New Mexico. J. Hydrol. 6, 194201. Dye, P. J. & Croke, B. F. W. 2003 Evaluation of streamflow predictions by the IHACRES rainfall-runoff model in two South African catchments. Environ.
Model. Soft. 18(89), 705712.

Hall, M. J. 1977 The effect of urbanization on storm runoff from two catchment areas in North London. Internat. Assoc. Hydrol. Sci. 123, 144152. Hansen, D. P., Ye, W., Jakeman, A. J., Cooke, R. & Sharma, P. 1996 Analysis of the effect of rainfall and streamflow data quality and catchment dynamics on streamflow prediction using the rainfall-runoff model IHACRES. Environ.
Soft. 11(13), 193202.

Hrissanthou, V., Mylopoulos, N., Tolikas, D. & Mylopoulos, Y. 2003 Simulation modeling of runoff, groundwater flow and sediment transport into Kastoria Lake, Greece. Water Res. Manag. 17(3), 223242. Jaber, J. O., Probert, S. D. & Badr, O. 1997 Energy and environmental issues for Jordan. Appl. Energ. 57(1), 45101. Jakeman, A. J. & Hornberger, G. M. 1993 How much complexity is warranted in a rainfall-runoff model. Water Res. Resear. 29(8), 26372649. Jakeman, A. J., Littlewood, I. G. & Whitehead, P .G. 1990 Computation of the instantaneous unit-hydrograph and identifiable component flows with application to 2 small upland catchments. J. Hydrol. 117(14), 275300. Kokkonen, T., Koivusalo, H. & Karvonen, T. 2001 A semi-distributed approach to rainfall-runoff modeling a case study in a snow affected catchment. Environ.
Model. Soft. 16(5), 481493.

71

Kokkonen, T. S., Jakeman, A. J., Young, P. C. & Koivusalo, H. J. 2003 Predicting daily flows in ungauged catchments: model regionalization from catchment descriptors at the Coweeta Hydrologic Laboratory, North Carolina. Hydrol.
Process. 17(11), 22192238.

Lane, L. J. & Renard, K. G. 1971 Evaluation of a basin-wide stochastic-model for ephemeral runoff from semi-arid watersheds. Transac. Asae 15(2), 280&. Lange, J., Liebundgut, C. & Schick, A. P. 2000 The importance of single events in arid zone rainfall-runoff modeling. Phys. Chem. Earth 25(78), 673677. Littlewood, I. G. 2003 Improved unit hydrograph identification for seven Welsh rivers: implications for estimating continuous streamflow at ungauged sites.
Hydrological Sciences Journal-Journal Des Sciences Hydrologiques 48(5),

743762. Lopezbermudez, F. 1990 Soil-erosion by water on the desertification of a semi-arid mediterranean fluvial basin the Segura Basin, Spain. Agri. Ecosys. Environ.
33(2), 129145.

McIntyre, N. & Al-Qurashi, A. 2009 Performance of ten rainfall-runoff models applied to an arid catchment in Oman. Environ. Model. Soft. 24(6), 726738. Merz, R. & Bloschl, G. 2004 Regionalisation of catchment model parameters. J.
Hydrol. 287(14), 95123.

Nash, J. E. 1959 Systematic determination of unit hydrograph parameters. J. Geophys.

Resear. 64(1), 111115.

Osborn, H. B. and Hickok, R. B. 1968 Variability of rainfall affecting runoff from a semi-arid rangeland watershed. Water Res. Resear. 4(1), 199&. Pilgrim, D. H., Chapman, T. G. & Doran, D. G. 1988 Problems of rainfall-runoff modeling in arid and semi-arid regions. Hydrol. Sci. J. 33(4), 379400.

72

Post, D. A. & Jakeman, A. J. 1999 Predicting the daily streamflow of ungauged catchments in SE Australia by regionalising the parameters of a lumped conceptual rainfall-runoff model. Ecolog. Model. 123(23), 91104. Rao, R. A., Delleur, J. W. & Sarma, P. B. S. 1971 Effects of urbanization on instantaneous unit hydrographs and on runoff. Transac. Amer. Geophys. Union
52(4), 198&.

Safriel, U. N. 2006 Dryland development, desertification and security in the Mediterranean. In Desertification in the Mediterranean Region. A Security
Issue (ed. Kepner, W. G., Rubio, J. L., Mouat, D. A. & Pedrazzini, F.),

Springer, 3, 227250. Schreider, S. Y., Jakeman, A. J., Falkland, A. & Knee, R. 1995 Streamflow prediction for the Queanbeyan River at Tinderry, Australia. Environ. Internat. 21(5), 545550. Sefton, C. E. M. & Howarth, S. M. 1998 Relationships between dynamic response characteristics and physical descriptors of catchments in England and Wales.
J. Hydrol. 211(14), 116.

Wheater, H. S., Sorooshian, S. & Sharma, K. D. 2008 Hydrological Modeling in Arid

and Semi-arid Areas, Cambridge University Press, Cambridge, 206 pp.

Whitehead, P., Wade, A. & Flynn, N. 2006 Water life and civilization project, second year report.

73

Appendix B
Abushandi, E. and Merkel, B.: Rainfall estimation over the Wadi Dhuliel arid catchment, Jordan from GSMaP_MVK+, Hydrol. Earth Syst. Sci. Discuss., 8, 1665-1704, doi:10.5194/hessd-8-1665-2011, 2011. (Accepted for discussion)

Eyad Abushandi, Broder Merkel

Technische Universitt Bergakdemie Freiberg, Geology Department Gustav-Zeuner-Str. 12, 09596 Freiberg/Germany, Office Tel.: +49 3731 39 3118 Email: [email protected]

Abstract:
The GSMaP_MVK+ (Global Satellite Mapping of Precipitation) dataset was used to evaluate the precipitation rates over the Wadi Dhuliel arid catchment in northeast Jordan for the period of January 2003 to March 2008. The scarcity of the ground rain gauge network alone did not adequately show the detailed structure of the rainfall distribution, independent form interpolation techniques used. This study combines GSMaP_MVK+ and ground rain gauges to produce accurate, high-resolution datasets. Three meteorological stations and six rain gauges were used to adjust and compare GSMaP_MVK+ estimates. Comparisons between GSMaP_MVK+ measurements and ground rain gauges records showed distinct regions where they correlate, as well as areas where GSMaP_MVK+ systematically over- and underestimated ground rain gauge records. A multiple linear regression (MLR) model was used to derive the relationship between rainfall and GSMaP_MVK+ in conjunction with temperature, relative humidity, and wind speed. The MLR equations were defined for the three meteorological stations. The best fit of MLR model for each station was chosen and

74

used to interpolate a multiscale temporal and spatial distribution. Results show that the rainfall distribution over the Wadi Dhuliel is characterized by clear west-east and north-south gradients. Estimates from the monthly MLR model were more reasonable than estimates obtained using daily data. The adjusted GSMaP_MVK+ performed well in capturing the spatial patterns of the rainfall at monthly and annual time scales while daily estimation showed some weakness in light and moderate storms.
Keywords: Rainfall, Arid Regions, IDW, GSMaP_MVK+, MLR

Introduction

Jordan has one of the worlds lowest levels of available water resources (WHO, 2011). Due to this scarcity, Jordanian scientists and government have taken an increasingly active role in studying and managing water within Jordan during the last decade. Around 91% of Jordan lies on arid and semi-arid ground which receive less than 200 mm of total annual rainfall (Figure.B 1). Precipitation occurs primarily as rainfall with relatively high intensity in limited range of space and time. Since precipitation is Jordans first source of water, it is important to investigate and analyze the rainfall behaviour. The rainfall distribution in Jordan varies with location mainly due to arid climatic conditions and topographic variations. Rainfall controls domestic and agricultural activities, especially in the rural area where the percentages of water use are 31% and 65% respectively (Aquastat-FAO, 2009). In comparison to other Middle Eastern countries, Jordan has the lowest magnitude of annual rainfall coincident with high evaporation rates. Rainfall is also the most important input parameter in rainfall runoff models (Beven 2001a; Croke and Jakeman 2008), groundwater recharge models (Abdulla and Al-Assa'd 2006; Merkel and Sperling 1993), climate change scenarios (Dolman and Gregory 1994) and hydro-chemical models (Brezonik and Stadelmann 2002). 75

Additionally, rainfall information is a critical component in efficient management of urban drainage systems (Vieux and Vieux 2005). Consequently, an accurate assessment of rainfall variability is essential to reduce models uncertainty in the input data of these models. Due to the arid climate, topographic variations, and a complicated land cover structure temporal and spatial rainfall distributions in Jordan are characterized by a high degree of variability. The annual rainfall magnitudes distinctly include a sharp west-east gradient from relatively wet west regions with about 600mm per year, to the Jordanian desert (Al-Badia), with rainfall less than 100mm per year.

Figure.B 1| Rainfall distribution in Jordan (JMWI)

The surface water resources and ground water recharges in the country depend on the magnitude of yearly rainfall. The total annual rainfall on Jordan is approximately 8500 106m3 (Abu-Zreig et al 2000). According to the Jordan Ministry of Water and Irrigation (JMWI), the majority of the rainfall is lost through evapotranspiration (87.9%), 8.5% is groundwater recharge, while the smallest portion is surface runoff (3.6%) (Figure.B 2A). These distributions are slightly different in drier regions of north east Jordan (Figure.B 2B).

76

Figure.B 2| The percentages of hydrologic water balance for Jordan (A) and

particularly the Wadi Dhuliel in northeast Jordan (B) (JMWI) Rainfall temporal analysis may include rainfall periodicity (Gajiccapka 1994), risk of drought (Akhtari et al 2009; Pal and Al-Tabbaa 2009; Wong et al 2010), chance of rain and frequency (Goldreich 1995), and time series analyses (Momani 2009). In contrast, spatial rainfall analysis focuses on the rainfall distribution within a watershed. Many different rainfall interpolation methods, such as Arithmetic Average, Isotheral method, and the Grid method are employed in current scientific literature. Thiessen polygons are the simplest interpolation method to estimate areal rainfall at a sample point (Thiessen 1911), These methods may not be the optimal to estimate the temporal and spatial rainfall changes in arid regions without additional information or techniques. In some cases, these methods were included in a hybrid approach that utilized other datasets and techniques in order to ensure output quality or to avoid rainfall observation scarcity. A relatively limited number of rainfall analysis techniques have been developed and modified for arid and semi-arid environments. Perhaps the earliest published research on rainfall magnitudes analysis in arid regions was performed by Winkwort (1967) in Australia and Osborn and Hickok (1968) in the United States. 77

Generally, the few published studies available from Jordan have tried to analyze rainfall characteristics in the entire country, rather than for individual catchments (Freiwan and Kadioglu 2008b; Tarawneh and Kadioglu 2003), even though many arid drainage basins might be smaller than 10 km2 (Pilgrim et al 1988). Two studies have focused on the techniques of optimizing the number and location of rain gauges (Manik and Sidle 2003; Tarawneh and Kadioglu 2003). Comprehensive surface hydrology studies including rainfall characteristics with respect to temporal and spatial variability have been carried out in Jordan for the last decade. Some of these studies examined the changes of rainfall temporal patterns only; other cases analyze both spatial and temporal patterns. In 2009, Momani analyzed the monthly rainfall temporal variation by applying ARIMA time series analysis to data recorded at the Amman airport. In order to achieve a proper rainfall forecast of his research, ARIMA model parameters were adjusted (Momani 2009). Smadi and Zghoul (2006) examined the recent rainfall temporal trends and fluctuations for three meteorological stations, Amman, Madaba, and Al-Mafraq. They observed a direct interrelationship between rainfall levels at these stations. Dahamsheh and Aksoy (2007) studied the structural characteristics of annual precipitation data for 13 meteorological stations distributed across Jordan and utilized the Isohyetal method to plot rainfall distribution. They employed a number of tests, such as consistency, randomness, best-fit distribution, and others in order to characterize the annual precipitation. There was no evidence of negative or positive precipitation trends at any station. However, these results can not be directly compared with previous studies.

78

Tarawneh and Kadiolgu (2003) selected seventeen meteorological stations corresponding to different climatic regions of Jordan in order to depict spatial monthly precipitation characteristics. The frequency amplitude, periodicity phase angle, and basic statistical parameters from the meteorological stations were calculated as steps of harmonic analysis of the precipitation. The results showed that the variance percentage of harmonic analysis is changing rapidly by moving to the east. According to the results of local studies water harvesting is one possible future solution to capture and store rainfall in Jordan (Abu-Zreig et al 2000; AbuAwwad and Shatanawi 1997; Oweis and Taimeh 1995). In order to achieve highest efficiencies, a thorough knowledge of rainfall distribution is essential. Spatial rainfall analysis requires a network of rain gauges or meteorological stations. The accuracy of spatial rainfall interpolation method depends on the density distribution and the distance between rainfall rain gauges. Frequently, rain gauge density is not sufficient in arid regions (Pilgrim et al 1988), leading to biased analyses of rainfall temporal and spatial distributions at the catchment scale. State-of-the-art techniques may solve this issue by matching precipitation data from ground-based rain gauges and high-resolution satellites in hybrid interpolation analysis. Recently, a number of global precipitation systems have been developed to meet scientific demand, such as PERSIANN (Sorooshian et al 2000), Global Precipitation Climatology Project (GPCP) (Adler et al 2000; Huffman et al 2001; Xie et al 2003), and Multi-satellite Precipitation Analysis (TMPA) (Huffman et al 2007). Global or near-global satellite datasets are important to identify temporal and spatial rainfall changes in arid regions. Global Satellite Mapping of Precipitation (GSMaP) is based on passive microwave radiometer data and has shown to be

79

effective for accurately estimating rainfall rate in mm/h (Ushio et al 2009). GSMaP combines precipitation retrievals from polar-orbiting satellites and cloud motion vectors derived from infrared images. GSMaP_MVK+ uses four different types of satellite sensors as shown in table.B 1 and an algorithm combining the CMORPH technique and Kalman filter (Tian et al 2010).

Table.B 1| Input datasets to produce GSMaP_MVK+ from four different types of satellite sensors
Input data Sensor GSMaP nearrealtime system GSMaP standard system

Passive microwave radiometer

TRMM/ The Tropical Rainfall Measuring Mission (TMI) Aqua/AMSR-E DMSP/SSMI(F13, 14, 15)

NASA/GSFC Real-time Version JAXA/EORC NOAA/NWS Globally-merged pixel-resolution data by JWA JMA Global Analysis (GANAL) Real-time Version JMA MGDSST

NASA/GSFC Standard Version JAXA/EORC Remote Sensing Systems Globally-merged pixel-resolution data by GSFC/DAAC JMA Global Analysis (GANAL) JMA MGDSST

GEO Infrared radiometer Atmospheric information Sea Surface Temperature

MTSAT, METEOSAT7/8,GOES-11/12 ---

---

The aim of this paper is to investigate rainfall characteristics of the Wadi Dhuliel catchment in northeastern Jordan by utilizing GSMaP data and ground based rain gauge data. Moreover, this study aims to develop a technique to adjust the GSMaP data by means of rain gauge data and standard interpolation techniques to perceive a good understanding of rainfall variability in arid regions, reduce the potential errors of rain gauge estimates, and produce improved catchment scale rainfall distribution maps.

80

2
2.1

Materials and methods

Study area description

The Wadi Dhuliel is an appropriate example of an arid catchment in which rainfall and rainfall intensity varies significantly both with with time and space. The total area of the Wadi Dhuliel is approximately 1985 km2 and is located in northeast of Jordan. Most of the catchment area belongs to the Al-Zarqa river basin. Around 10% of the upper part of Wadi Dhuliel catchment passes over the Syrian border (Figure.B 3). The altitude in the area is characterized by a very gentle undulating topography varying between 512m in the southwest to 1400m in the north.

Figure.B 3| Study area location map of Al-Zarqa basin including the sub-

basins Seil Al-Zarqa and Wadi Dhuliel.

81

Rainfall temporal magnitude in the Wadi Dhuliel tends to vary markedly from year to year with an irregular distribution over the year. As an example of the extreme yearly variability in Wadi Dhuliel one rain gauge measured the annual rainfall to be 275.7, 93.1, 111.1, 230.4, 194.8, 63.1, and 209.5 mm over seven years. In one single day, a 62 mm rainfall event occurred, though the total annual rainfall in the same year was 100 mm (Sukhnah rain gauge). These kinds of rainfall events can easily generate significant surface runoff, resulting in severe soil erosion. Weather behaviour and topographical aspects play important roles in this variation. The region has essentially an arid climate with cold, rainy winter and a hot, dry summer. The average monthly rainfall showed that around 73% of the annual rainfall occurs during November, December, January, and February (Figure.B 4).

Figure.B 4| Monthly average rainfall (1976-2005)

Almafraq station has the highest rainfall magnitudes per annum with 158mm, Qasr Al-Hallabat station has the lowest rainfall with 79.2mm. Overall, the annual rainfall is around 123mm on average. In addition, the lowest temperatures are also during the winter months, with an average annual temperature (1976-2005) of 16.8 C (Figure.B 5). 82

Figure.B 5| Monthly average of the minimum and maximum

temperature (1976-2005)
2.2 Data and method of data analysis

The datasets used in this work included ground rainfall data of nine gauging stations at daily, monthly, and annual time steps between January 2003 and March 2008. A complementary Global Satellite Mapping of Precipitation dataset, currently known as GSMaP_MVK+ version 4.8.4 (short for GSMaP moving vector with Kalman filter method), was also examined.

Figure.B 6| Research process flowchart

83

The rain gauges dataset (Table.B 2) was gathered from the Surface Water Resources Unit at the Jordan Ministry of Water and Irrigation (JMWI), and the Jordan Meteorological Department (JMD). Eight rain gauges are distributed in and near to Wadi Dhuliel I, and one station is located in Wadi Dhuliel II (Figure.B 7). Almafraq meteorological station has records only until 2005. However, only these nine stations produced a sufficient length of reliable data appropriate for this study.

Table.B 2| Information about meteorological and rain gauging stations utilized for the study
Station code on the map (Figures.B 3 and 7) A B C D E F G H I
*

Station code (JMWI) AL0058 AL0059 AL0048 AL0055 AL0049 AL0054 AL0066 No code No code

Station name

Altitude [above M.S.L. in meters] 843 670 600 580 590 566 564 675 675

Data type

Mean annual rainfall [mm] 109.3 110 123.9 130.3 72.4 135.3 131.9 158 143

Sabha and Subhiyeh Um-Jimal* Al-Khaldiya Wadi Dhuliel Nursery Qasr Al-hallabat Hashimiya Khirebit Es Samra Evap. St.* Almfraq Mafraq_60*

Monthly and daily Monthly and daily Monthly and daily Monthly and daily Monthly and daily Monthly and daily Monthly and daily Monthly Hourly

Meteorological station

The Global Satellite Mapping of Precipitation (GSMaP) project started in 2002 with support of the Japan Science and Technology Agency (Ushio et al 2009). A frame from 31.95oN - 32.55oN and 36.15oE 36.85oE was extracted from the GSMaP_MVK+ dataset to cover the entire Wadi Dhuliel catchment area with 64 knots (8 8) and a spatial resolution of 10.8 km (Figure.B 7). Based on the altitudes,

84

rainfall magnitudes, and land cover characteristics, the area was divided into two distinct sub-catchments Wadi Dhuliel I and Wadi Dhuliel II.

Figure.B 7| The GSMaP_MVK+ pixels distribution around and over

Wadi Dhuliel catchment (Rainfall mm/month, [January 2003])

Rainfall ground dataset was based upon the acquisition period of GSMaP_MVK+ data, from January 2003 to April 2008. In order to assess at which time scale the GSMaP_MVK+ estimates have sufficient match, the daily datasets from the GSMaP_MVK+ and ground rain gauging station were aggregated to monthly and annual records. For the comparison between ground rain gauges and GSMaP_MVK+ datasets values of all ground rain gauge station were calculated from the four neighbouring GSMaP_MVK+ knots using inverse distance weighting (IDW) interpolation method.

85

2.3

Satellite measurements adjustment with rain gauges

Since the GSMaP_MVK+ algorithm has been developed for precipitation over tropical and subtropical regions (Iwasaki 2009); ground observations are required to adjust the satellite information. Furthermore, the input data to the GSMaP_MVK+ is based upon brightness temperature and cloud microphysical properties, and therefore gives relatively indirect information about rainfall rate. This adjustment process is needed to tune the residuals between local observation datasets and GSMaP_MVK+ estimates. The adjustment process was based on remapping GSMaP_MVK+ pixel values with respect to rain gauge observations. Datasets from three meteorological stations were used to adjust GSMaP_MVK+ dataset. Khirebit Es Samra and Um-Jimal meteorological stations provide monthly rainfall, temperature, and wind speed data sets from 2003 and 2008, while Mafraq_60 meteorological station provides 1-hourly rainfall, temperature, and relative humidity records for the period between 2004 and 2006. Unfortunately, some hourly records are missing from Mafraq_60

meteorological station. As observed from Mafraq_60 station, most of the rain events are related to low temperature and high relative humidity (Figure.B 8 and table 3). Furthermore, a significant correlation between rainfall, temperature and relative humidity (RH) was observed. The Spearman correlation coefficient (rho) between hourly temperature and rainfall rate is -0.28 (two tailed P =0.48), while Spearmans rho is 0.089 (two tailed P=0.026) between rainfall rate and RH. Hourly wind speed records have also a positive correlation coefficient with rainfall records but not significant (rho =0.122, two tailed P =0.002). However, some anomalous satellite pixel values are detected and skipped from the adjustment process.

86

Figure.B 8| Comparison between hourly rainfall rate with air

temperature, relative humidity and wind speed from Mafraq_60 meteorological station (2004-2006) The next step towards adjustment is to aggregate 1-hourly dataset into daily, monthly, and annual datasets. For this the daily rainfall rates have been categorized into three groups: (i) Light 0.1-1.0 mm/day (ii) Moderate 1.1-5.0 mm/day, and (iii) Heavy >5.0 mm/day. Zero values from both ground gauges and GSMaP_MVK+ were excluded. Consequently, GSMaP_MVK+ pixel estimates were compared to daily and

87

monthly ground rain, temperature, wind speed, and relative humidity (Figures.B 9 and 10). The comparison shows three groups: 1. GSMaP_MVK+ estimates matched the rain ground records rather fairly 2. GSMaP_MVK+ values are underestimates 3. GSMaP_MVK+ values are overestimates

Figure.B 9| The daily rainfall rates at the ground and their estimates by

GSMaP_MVK+ from Mafraq_60 (2004-2006)

88

Figure.B 10| The monthly rainfall rates monitored at ground and their

estimates by GSMaP_MVK+ from Um-Jimal metrological station (Left) and Khirebit Es Samra metrological station (Right) (2003-2008)

In order to categorize monthly rain events into groups of similar trajectories K-means clustering was applied (Tables.B 3 and 4). The aim of this categorization was to assess the effect of external variables on rainfall rates to GSMaP_MVK estimates, air temperature, and wind speed records. The events with totals less than 2mm/month were excluded.

89

Table.B 3| K-means cluster analysis of monthly air temperature, wind speed, and GSMaP_MVK+ for Um-Jimal metrological station (2003-2008)
Variable Average Temperature [C] Wind Speed [km/h] Cluster No 1* Cluster No 2** Std. Deviation Std. Deviation [Cluster1 ] [Cluster2 ]

11.3 9.3

10.0 5.54

3.56 5.54 16.6

6.4 1.6 48.8

GSMaP_MVK+ 23.3 105.9 [mm/month] * The number of cases in cluster number 1 is 22 ** The number of cases in cluster number 2 is 2

Table.B 4| K-means cluster analysis of monthly air temperature, wind speed, and GSMaP_MVK+ Khirebit Es Samra metrological station (2003-2008)
Variable Average Temperature [C] Wind Speed [km/h] Cluster No 1* Cluster No 2** Std. Deviation Std. Deviation [Cluster1 ] [Cluster2 ]

10.77 2.5

13.3 2.7

3.52 1.17 12.92

2.87 0.9 20.1

GSMaP_MVK+ 86.45 15.6 [mm/month] * The number of cases in cluster number 1 is 6 ** The number of cases in cluster number 2 is 19

Multiple linear regressions (MLR) method was carried out to estimate convective and stratiform rain rates from GSMaP_MVK+ among temperature, relative humidity, and wind speed of daily and monthly records. The relationship between the ground rainfall records and the explanatory variables is represented for each station by the following equations:

90

1. The MLR model for Mafraq_60 meteorological station daily records:

Rf Light = GSMaP 0.012 + Temp 0.023 + RH 0.002 C1 Rf Moderate = GSMaP 0.027 Temp 0.068 + RH 0.01 + C 2 Rf Heavy = GSMaP 1.7 + Temp 1.03 + RH 0.99 C3

(Equation 1) (Equation 2)

(Equation 3)

Where Rf Light stands for the rainfall rate between 0.1-1.0 mm/day, Rf Moderate is the rainfall between rate between 1.1- 5.0 mm/day, Rf Heavy is the rainfall rate more than 5.0 mm/day, GSMaP is the GSMaP_MVK+ estimates version 4.8.4 recorded in mm/day, Temp is the temperature records in Celsius degree, RH is the relative humidity in percentage, C1 is the Rf Light constant and equal to 0.164 , C 2 is the
Rf Moderate constant and equal to 4.46, and C 3 is the Rf Heavy constant and equal to 71.8.

2. The MLR model for Um-Jimal meteorological station monthly records:

Rf cluster1 = GSMaP 0.29 Temp 1.85 + WS 1.008 + Ccluster1 Rf cluster 2 = GSMaP 3.53 + C cluster 2

(Equation 4) (Equation 5)

Where Rf cluster1 stands for the rainfall rate in mm/month for the first group of cluster, Rf cluster 2 is the rainfall rate in mm/month for the second cluster, and C cluster1 is the first cluster constant and equal to 38.8, C cluster 2 is the second cluster constant and equal to 78.42, and WS is wind speed in km/h.

91

3. Monthly records of Es Samra meteorological station:

Rf cluster1 = GSMaP 1.03 + Temp 1.86 WS 5.13 Ccluster1 Rf cluster 2 = GSMaP 0.64 Temp 2.77 + WS 5.12 + Ccluster 2

(Equation 6) (Equation 7)

Where C cluster1 is equal to 44.2 and C cluster 2 is equal to 56.1

Then, the MLR equations were chosen from all these combination to adjust GSMaP_MVK+ estimates:
GSMaP _ MVK recalibrated = GSMaP _ MVK original (GSMaP _ MVK original RE )

(Equation 8) RE is the relative error and can be determined by using the following equation:
RE % = ( Rf recalibrated GSMaP _ MVK original ) / GSMaP _ MVK original

(Equation 9)

In order to measure the reliability of simulated values the bias calculation assesses the average difference between GSMaP_MVK+ and ground rain gauges:
BIAS = 1 n (GSMaP _ MVK recalibrated Gauge) n i =1

Where GSMaP _ MVK recalibrated is the GSMaP_MVK+ estimates after the adjustment process at the station location, Gauge is the observed rainfall ground. N is the number of elements.

92

Results
Using the available weather records between 2003 and 2008 from nine

meteorological and rain gauge stations over the Wadi Dhuliel complementary with GSMaP_MVK+ rainfall data showed a complex rainfall pattern in the Wadi Dhuliel. The evaluation of daily and monthly GSMaP_MVK+ datasets exhibited good performance in capturing relative values of rainfall pattern but poor results with respect to estimating the absolute values of the rainfall. The comparison of daily and monthly GSMaP_MVK+ and ground records showed significant under- and overestimations in both spatial and temporal distributions. Separate from cases where GSMaP_MVK+ and ground records are correlated, in general GSMaP_MVK+ records showed overestimation. Daily records of the GSMaP_MVK+ are showing 84% overestimation while in monthly records it is 59%. Most of the annual rainfall magnitudes of GSMaP_MVK+ were overestimates (85.7%); only the year 2003 exhibited correlation (Table.B 5).

93

Table.B 5| Annual rainfall of GSMaP_MVK+ compared to 8 ground rain gauge stations Data Type Date Hashimiya Khirebit Es-Samra W. Dhuliel Nursery Um El-Jumal GSMaP 114.02 136.44 125.71 162.08 2003 Ground Record 184.2 194 138.9 172.8 GSMaP 365.44 400.99 396.99 597.43 2004 G Record 114.6 130 85.4 74.2 GSMaP 232.68 256.72 231.26 455.99 2005 Ground Records 122.9 124 86.2 105 GSMaP 321.61 334.07 308.5 367.56 2006 Ground Records 102.8 86.8 97.5 79.2 GSMaP 10.04 10.4 9.65 21.58 Dec-07* Ground Records 14 12.6 8.5 23 GSMaP 112.35 109.03 123.22 181.09 2008 Ground Records 77.7 77 46.8 83 Data Type Date Khaldiya Sabha and Subhiyeh Qasr El-Hallabat Al-Mafraq GSMaP 127.37 161.7 80.43 162.95 2003 Ground Record 143.9 146.9 40.8 54.6 GSMaP 374.49 567.42 295.48 385.96 2004 G Record 91.5 86.7 74.7 105.7 GSMaP 248.58 412.07 153.67 466.42 2005 Ground Records 100.7 97.1 69.7 123.7 GSMaP 309.86 361.51 311.34 396.41 2006 Ground Records 99.3 96.8 47 NA GSMaP 11.46 20.8 10.24 20.5 Dec-07 Ground Records 8.5 16.5 6 NA GSMaP 124.68 164.18 120.67 122.7 2008** Ground Records 62 105.1 30.4 NA * The available month from 2007 is December only
**

Jan, Feb, and March in 2008 have no error estimates

In order to match GSMaP_MVK+ values in all cases within some acceptable error, an adjustment was performed based on ground data based on multiple linear regressions. For daily time step, MLR equations were calculated for Mafraq_60 meteorological station and derived from temperature and relative humidity records as well as from GSMaP_MVK+ daily estimates. Clustering rain events into light, moderate, and heavy storm helped to reduce the relative error. For monthly time step, explanatory variables used to develop MLR equations were calculated from two meteorological stations: Um-Jimal and Khirebit Es Samra. Here, the MLR equations

94

derived from temperature and wind speed records as well as from GSMaP_MVK+ monthly estimates. Unfortunately monthly relative humidity is not available for these two stations. If the relationship between variables is not clear, clustering of monthly rainfall to groups with respect to the variation of temperature, wind speed, and GSMaP_MVK+ estimates is a primary step to develop MLR models. The results of daily and monthly rainfall rates including underand overestimates corresponding to each

meteorological station after the calibration process are shown in Figures.B 11 and 12. The results showed good agreement between adjusted rainfall rates with ground station observations. The Spearmans correlation coefficient between adjusted and observed values for daily records shows significant correlation. The heavy storm events correlation coefficient was 0.75 (P =0.084), while for light and moderate storm events rho was 0.62 (two tailed P =0.008) and 0.66 (two tailed P =0.071), respectively. This may reflect the effect of extreme rain rates on Spearmans correlation coefficient.

95

Figure.B 11| Comparison between daily adjusted rainfall rates from

Temperature, relative humidity, and GSMaP_MVK+ records with ground rainfall rates obtained from Mafraq_60 meteorological station (2004-2006)

96

Figure.B 12| Comparison between monthly adjusted rainfall rates from

Temperature, wind speed, and GSMaP_MVK+ records with ground rainfall rates obtained from Um-Jimal metrological station (Left) and Khirebit Es Samra metrological station (Right) (2003-2008)

Spatial rainfall analysis was based on Inverse Distance Weighting (IDW) interpolation method. Daily results included one meteorological station (Mafraq_60) and seven rain gauges. The MLR model was carried out for Mafraq_60 meteorological station and extended to the daily GSMaP_MVK+ pixel values (equations 8 and 9). Adjusted daily GSMaP_MVK+ performed well in capturing the spatial patterns of the rainfall distribution, and showing more details especially on

97

extreme rainfall events, while some weakness in light and moderate storms spatial distributions (Figure.B 13). The MLR model was computed for monthly records acquired from Khirebit Es Samra and Um-Jimal meteorological stations for the time between 2003 and 2008. For ground interpolation, six other rain gauges were also used. The adjustment of GSMaP_MVK+ was primarily based on the average of equations 8 and 9 calculated from both stations. The MLR model monthly rainfall estimates were found to be more reasonable than estimates obtained using daily MLR. The evaluation of spatial patterns shows that monthly GSMaP_MVK+ does well in capturing the topographic effect on precipitation distribution pattern, in particular for the west-east and north-south precipitation gradients (Figure.B 14). A key outcome of the spatial and temporal analyses is the advantage of aggregating the fine scale data to coarser resolution (Figure.B 15).

98

Figure.B 13| Spatial comparison of daily rainfall between re-calibrated

GSMaP_MVK+ estimates (Left) and eight ground rainfall station records (right) using IDW method, light storm in 24.12.2004, moderate storm in 09.03.2005, and heavy storm in 06.02.2005

99

Figure.B 14| Spatial comparison of mean monthly rainfall between adjusted

GSMaP_MVK+ estimates (Left) and eight ground rainfall stations records (right) using IDW method, an example from January 2003

Figure.B 15| Spatial comparison of average annual rainfall between adjusted

GSMaP_MVK+ estimates (Left) and eight ground rainfall stations records (right) using IDW, an example from 2004

100

Discussion
Drawing conclusions from two disparate datasets may improve the quality of

the combined data. Due to the sparseness of the rain gauge networks, interpolated data often are biased by the interpolation algorithms. In some cases, the ground rainfall gauges reveal slight homogeneity of rainfall magnitudes but the spatial distributions are, in general, heterogeneous. Furthermore, the number of stations and the length of historical records affect both spatial and temporal correlation structures. The results of this study are in agreement with previous works which showed that the characteristic of rainfall in arid catchment varies in space and time (Abu-Zreig et al 2000; Lange et al 2000; Pilgrim et al 1988). Overall, GSMaP_MVK+ showed the best performance in comparison with other satellite products. This conclusion has been proven by several authors (Dinku et al., 2010a; Dinku et al., 2010b; Iwasaki, 2009; Ushio et al., 2009). The results of (Dinku et al 2010a; Dinku et al 2010b; Iwasaki 2009; Ushio et al 2009) motivated us to use the GSMaP_MVK+ in our study. Results of GSMaP_MVK+ were crosschecked against nine rain gauges observations assuming to represent reasonable and reliable point data. The compatibility between

GSMaP_MVK+ and the ground rain gauges was limited to specific months. The overor underestimation of the GSMaP_MVK+ in estimating rainfall in arid regions may be influenced by the following factors: a) The sensors detected the rainfall aloft, meaning the rain may have evaporated before reaching the ground (Dinku et al 2010a; Rosenfeld and Mintz 1988). b) The moving vector Kalman filter algorithm was developed for precipitation over the tropical and sub-tropical regions (Iwasaki 2009) using IR data as a means to move the precipitation estimates from microwave observation during

101

periods when microwave data are not available. Obviously this estimate does not work always properly in arid areas. c) The available GSMaP_MVK+ product was originally calibrated using ground based radar data located in tropical and sub-tropical regions of Japan, which may have different weather regime or covered by different cloud systems (Petty 2001) than in arid regions. d) An abrupt change in wind speed or wind direction below the cloud may have affected the rainfall area. A study conducted in Israel and Jordan by Sharon (1978) showed that the expansion of rainfall area may not be fully represented by point measurements. An increase of 10 km/h would constitute 12-15% of total rainfall. e) The rainfall duration varied from storm to storm. The rainfall storm over the study area was characterized by high rainfall intensity (Figure.B 16). This, however, might have occurred at the time when no satellite was overhead. f) Desert dust and other aerosols may suppress rainfall and alter cloud microphysical properties (Han et al 2008; Rosenfeld et al 2001). The desert dust above and in the cloud could have distorted the satellite measurements (Rosenfeld et al 2001). However, most of the previous related studies were usually based on homogenous water cloud models(Schutgens and Roebeling 2009). Generally, there is no doubt that remote sensing capability to monitor rainfall is still under development and has -in particular- some weaknesses related to arid regions application. However, there are on-going projects to improve the correlation between Satellite based rainfall retrieval algorithms and the real rainfall magnitudes (e.g. The Wet-Net Precipitation Inter-comparison Project, and the NASA GPM

102

Project). As previously explained, it was seen that GSMaP_MVK+ over the Wadi Dhuliel arid catchment exhibited poor performance in compare with ground data. Therefore, developing a new framework to re-adjust the GSMaP_MVK+ data by means of ground data and standard interpolation techniques was logically necessary. This new concept has caught our attention in order to achieve a reasonable representation of the true rainfall distribution.

Figure.B 16| The duration of a single storm event recorded at 10 minute

intervals (25.12.2008) Precipitation in this area is very spotty in both time and space. The datasets in figure.B 16 were obtained from our station which was newly installed in the study area. In addition, global warming increases the rainfall variability and change the rain form at global and regional scales, moreover, might cause extreme events. A study made by Abushandi and Merkel (2011) showed that the study area received unusual snowstorm in 1990. This snowstorm might give an indicator of global warming crossing the Wadi Dhuliel arid catchment. Since rainfall is highly fluctuated in arid 103

regions, the rising of temperature gradients is pretty much obvious than the rainfall variability. However, studying the impact of climate change on rainfall variability from space requires longer period datasets. We are fully aware of the uncertainties associated with satellite derived rainfall maps and the original calibration from ground radar. Therefore an adjustment process for GSMaP_MVK+ results was needed to achieve better match with ground observations in arid regions. Multiple linear regression analysis proved to be an appropriate technique using additional weather data. As expected, rainfall events showed higher intensities in the western parts, while eastern parts are characterized by lower rainfall rates. In some cases the rainfall distribution did not show the west-east gradient, as can be seen in the Appendix. This may be related to climatic and topographic variations. The high-resolution GSMaP_MVK+ dataset allowed us to evaluate and estimate the amount of rainfall in regions where no ground rainfall stations were available. Thus, the gridding interpolation method provided a qualitative view of the rainfall distribution. However, it is important to note that the interpolation technique explicitly derived new spatial values based on the number of present rain gauges, and, if the number of the gauges is limited, the unknown points may not be interpolated properly. Allowance for other weather variables such as radiation, evaporation, and would improve the accuracy of Global Satellite Mapping of Precipitation estimates. Furthermore, employing other satellite and aircraft observation for retrieving clouds properties may enhance our understanding of the microphysical impact of aerosols on water clouds. However, the quality of this rainfall analysis will be affected by paucity of data in the region.

104

In this study, an attempt was made to apply a multiple linear regression (MLR) model to derive the relationship between rainfall and GSMaP_MVK+ in conjunction with temperature, relative humidity, and wind speed. The application of MLR seems to do consistently well under varying time scales. Since there is a very limited number of studies on using MLR in arid region, these results will confidently let us apply the MLR model in further research. In a further research, a new framework of hydrological models application in arid catchment will be developed by integrating re-adjusted satellite derived rainfall dataset (GSMaP_MVK+) as an input dataset.

Conclusion
The climate in the Wadi Dhuliel area is characterized by high rainfall

variability. Hence, it is difficult to estimate the spatial rainfall variability by a simple gridding method. Rainfall records from different rain gauges showed a complex rainfall regime in the area. Rainfall distribution in the Wadi Dhuliel varies with location mainly due to topographic variations as one move from semi-arid to arid regions. A Global Satellite Mapping of Precipitation dataset, currently known as GSMaP_MVK+ Version 4.8.4, was compared with eight rain gauge stations at monthly and annual time steps. The performance of GSMaP_MVK+ over arid regions in general and over the Wadi Dhuliel arid catchment in particular is limited and requires a re-adjustment process. The results showed how topographic variation can influence the rainfall distribution, especially in the northern part of the catchment. Higher rainfall rates in the western parts and the lower rainfall rates in eastern parts may explain the change in climate from arid area to desert area. Moreover,

105

aggregating hourly rain rate into coarser time step, daily and monthly, will contribute to more accurate rain estimation.

Acknowledgment
The authors are thankful to Jordan Ministry of Water and Irrigation, particularly the Surface Water Unit, and the Jordan Meteorological Department for providing the necessary datasets. Special thanks go to all members of the GSMaP project for providing the GSMaP dataset.

References
Abdulla, F., and T. Al-Assa'd, 2006, Modeling of groundwater flow for Mujib aquifer, Jordan: Journal of Earth System Science, v. 115, p. 289-297. Abu-Zreig, M., M. Attom, and N. Hamasha, 2000, Rainfall harvesting using sand ditches in Jordan: Agricultural Water Management, v. 46, p. 183-192. AbuAwwad, A. M., and M. R. Shatanawi, 1997, Water harvesting and infiltration in arid areas affected by surface crust: examples from Jordan: Journal of Arid Environments, v. 37, p. 443-452. Abushandi, E., and B. Merkel, 2011, Application of IHACRES rainfall-runoff model into the Wadi Dhuliel arid catchment: Journal of Water and Climate Change, v. 2, p. 5671. Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and E. J. Nelkin, 2000, Tropical rainfall distributions determined using TRMM combined with other satellite and rain gauge information: Journal of Applied Meteorology, v. 39, p. 20072023. Akhtari, R., S. Morid, M. H. Mahdian, and V. Smakhtin, 2009, Assessment of areal interpolation methods for spatial analysis of SPI and EDI drought indices: International Journal of Climatology, v. 29, p. 135-145. Aquastat-FAO, 2009, Water Report 34.

106

Beven, K., 2001, On fire and rain (or predicting the effects of change): Hydrological Processes, v. 15, p. 1397-1399. Brezonik, P. L., and T. H. Stadelmann, 2002, Analysis and predictive models of stormwater runoff volumes, loads, and pollutant concentrations from watersheds in the Twin Cities metropolitan area, Minnesota, USA: Water Research, v. 36, p. 1743-1757. Croke, B. F. W., and A. J. Jakeman, 2008, Use of the IHACRES rainfall-runoff model in arid and semi arid regions, in H. S. Wheater, Sorooshian, S., Sharma, K.D., ed., Hydrological Modelling in Arid and Semi-arid Areas: Cambridge, Cambridge University Press, p. 41-48. Dahamsheh, A., and H. Aksoy, 2007, Structural characteristics of annual precipitation data in Jordan: Theoretical and Applied Climatology, v. 88, p. 201-212. Dinku, T., P. Ceccato, K. Cressman, and S. J. Connor, 2010a, Evaluating detection skills of satellite rainfall estimates over Desert Locust Recession regions: Journal of Applied Meteorology and Climatology, v. 49, p. 1322-1332. Dinku, T., F. Ruiz, S. J. Connor, and P. Ceccato, 2010b, Validation and intercomparison of satellite rainfall estimates over Colombia: Journal of Applied Meteorology and Climatology, v. 49, p. 1004-1014. Dolman, A. J., and D. Gregory, 1994, The Parametrization of rainfall interception in Gcms - Reply: Quarterly Journal of the Royal Meteorological Society, v. 120, p. 735-736. Freiwan, M., and M. Kadioglu, 2008, Spatial and temporal analysis of climatological data in Jordan: International Journal of Climatology, v. 28, p. 521-535. Gajiccapka, M., 1994, Periodicity of Annual precipitation in different climate regions of Croatia: Theoretical and Applied Climatology, v. 49, p. 213-216. Goldreich, Y., 1995, Temporal variations of rainfall in Israel: Climate Research, v. 5, p. 167-179. Han, Y., X. Dai, X. Fang, Y. Chen, and F. Kang, 2008, Dust aerosols: A possible accelerant for an increasingly arid climate in North China: Journal of Arid Environments, v. 72, p. 1476-1489.

107

Huffman, G. J., R. F. Adler, D. T. Bolvin, G. J. Gu, E. J. Nelkin, K. P. Bowman, Y. Hong, E. F. Stocker, and D. B. Wolff, 2007, The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales: Journal of Hydrometeorology, v. 8, p. 38-55. Huffman, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce, B. McGavock, and J. Susskind, 2001, Global precipitation at one-degree daily resolution from multisatellite observations: Journal of Hydrometeorology, v. 2, p. 36-50. Iwasaki, H., 2009, NDVI prediction over Mongolian grassland using GSMaP precipitation data and JRA-25/JCDAS temperature data: Journal of Arid Environments, v. 73, p. 557-562. Lange, J., C. Liebundgut, and A. P. Schick, 2000, The importance of single events in arid zone rainfall-runoff modelling: Physics and Chemistry of the Earth Part B-Hydrology Oceans and Atmosphere, v. 25, p. 673-677. Manik, T., and R. Sidle, 2003, Rainfall spatial distribution in Sumber Jaya Watershed, Lampung, Indonesia. Merkel, B., and B. Sperling, 1993, Raster based modelling of watersheds and flow accumulation: Application of Geographic Information Systems in Hydrology and Water Resources, p. 193-199. Momani, N., 2009, Time series analysis model for rainfall data in Jordan: case study for using time series analysis: American Journal of Environmental Sciences, v. 5, p. 599-604. Osborn, H. B., and R. B. Hickok, 1968, Variability of rainfall affecting runoff from a semiarid rangeland watershed: Water Resources Research, v. 4, p. 199-203. Oweis, T. Y., and A. Y. Taimeh, 1995, Evaluation of a small basin water-harvesting system in the arid region of Jordan: Water Resources Management, v. 10, p. 21-34. Pal, I., and A. Al-Tabbaa, 2009, Trends in seasonal precipitation extremes - An indicator of 'climate change' in Kerala, India: Journal of Hydrology, v. 367, p. 62-69.

108

Petty, G. W., 2001, Physical and microwave radiative properties of precipitating clouds. Part II: A parametric 1D rain-cloud model for use in microwave radiative transfer simulations: Journal of Applied Meteorology, v. 40, p. 21152129. Pilgrim, D. H., T. G. Chapman, and D. G. Doran, 1988, Problems of rainfall-runoff modeling in arid and semiarid regions: Hydrological Sciences Journal-Journal Des Sciences Hydrologiques, v. 33, p. 379-400. Rosenfeld, D., and Y. Mintz, 1988, Evaporation of rain falling from convective clouds as derived from radar measurements: Journal of Applied Meteorology, v. 27, p. 209-215. Rosenfeld, D., Y. Rudich, and R. Lahav, 2001, Desert dust suppressing precipitation: A possible desertification feedback loop: Proceedings of the National Academy of Sciences of the United States of America, v. 98, p. 5975-5980. Schutgens, N. A. J., and R. A. Roebeling, 2009, Validating the validation: the influence of liquid water distribution in clouds on the intercomparison of satellite and surface observations: Journal of Atmospheric and Oceanic Technology, v. 26, p. 1457-1474. Sharon, D., 1978, Rainfall fields in Israel and Jordan and effect of cloud seeding on them: Journal of Applied Meteorology, v. 17, p. 40-48. Smadi, M., and A. Zghoul, 2006, A sudden change in rainfall characteristics in Amman, Jordan during the mid 1950s: American Journal of Environmental Sciences, v. 2, p. 84-91. Sorooshian, S., K. L. Hsu, X. Gao, H. V. Gupta, B. Imam, and D. Braithwaite, 2000, Evaluation of PERSIANN system satellite-based estimates of tropical rainfall: Bulletin of the American Meteorological Society, v. 81, p. 2035-2046. Tarawneh, Q., and M. Kadioglu, 2003, An analysis of precipitation climatology in Jordan: Theoretical and Applied Climatology, v. 74, p. 123-136. Thiessen, A. H., 1911, Precipitation averages for large areas: Monthly Weather Review, v. 39(7), p. 1082-1084.

109

Tian, Y. D., C. D. Peters-Lidard, R. F. Adler, T. Kubota, and T. Ushio, 2010, Evaluation of GSMaP precipitation estimates over the contiguous United States: Journal of Hydrometeorology, v. 11, p. 566-574. Ushio, T., K. Sasashige, T. Kubota, S. Shige, K. Okamoto, K. Aonashi, T. Inoue, N. Takahashi, T. Iguchi, M. Kachi, R. Oki, T. Morimoto, and Z. I. Kawasaki, 2009, A Kalman Filter approach to the Global Satellite Mapping of Precipitation (GSMaP) from combined passive microwave and infrared radiometric data: Journal of the Meteorological Society of Japan, v. 87, p. 137151. Vieux, B. E., and J. E. Vieux, 2005, Statistical evaluation of a radar rainfall system for sewer system management: Atmospheric Research, v. 77, p. 322-336. WHO, 2011, Pilot projects, Jordan: water is life, action at country level. Winkwort.Re, 1967, Composition of several arid Spinifex Grasslands of Central Australia in relation to rainfall soil water relations and nutrients: Australian Journal of Botany, v. 15, p. 107-130. Wong, G., M. F. Lambert, M. Leonard, and A. V. Metcalfe, 2010, Drought analysis using Trivariate Copulas conditional on climatic states: Journal of Hydrologic Engineering, v. 15, p. 129-141. Xie, P. P., J. E. Janowiak, P. A. Arkin, R. Adler, A. Gruber, R. Ferraro, G. J. Huffman, and S. Curtis, 2003, GPCP Pentad precipitation analyses: An experimental dataset based on gauge observations and satellite estimates: Journal of Climate, v. 16, p. 2197-2214.

110

Appendix C
Abushandi E., Merkel B. (2011) Modelling rainfall runoff relations using HEC-HMS and IHACRES for a single rain event in arid region, Jordan. Water Resources Management Journal (Ready for submission)

Eyad Abushandi, Broder Merkel

Institute for Hydrogeology, Technical University of Freiberg Gustav-Zeuner-Str. 12, 09596 Freiberg/Germany
Office Tel.: +49 3731 39 3118 Email: [email protected]

Abstract:
The HEC-HMS and IHACRES rainfall runoff models were applied to simulate a single streamflow event in the Wadi Dhuliel arid catchment occurring on 30./31.01.2008. The HEC-HMS model application was using the HEC-GeoHMS extension in ArcView 3.3. Streamflow estimation was performed on the basis of hourly scale. The aim of this study was to develop a new framework of rainfall-runoff model applications in arid catchment by integrating re-adjusted satellite derived rainfall dataset (GSMaP_MVK+) to determine the location of the rainfall storm. Each model has its own input data. HEC-HMS input data include soil type, land use/land cover , and slope. IHACRES input data sets include hourly rainfall and temperature. The model was calibrated and validated using observed stream flow data set collected at Al-Zaatari discharge station. The performance of IHACRES showed some weaknesses, while the flow comparison between the calibrated streamflow results fits well with the observed streamflow data in HEC-HMS model performance. The NashSutcliffe efficiency (Ef) for both models was 0.51, and 0.88 respectively. The application of HEC-HMS model is considered to be satisfactory.
Keywords: HEC-HMS, GIS, Arid regoin, Jordan, Wadi Dhuliel SCS-CN method;

111

Introduction
Poor availability of hydrologic studies for the Wadi Dhuliel arid catchment in the northern Jordan leads to miss planning and inadequate water recourses management strategies not only at the catchment scale but also for the entire Al Zarqa Basin. Implementation of storm water management is essential in the Wadi Dhuliel to increase water availability especially for agriculture and livestock sectors. However, most of the existing water monitoring projects in Jordan are evaluated as lacking comprehensive hydrologic data or accessibility. Based on this fact, there is a growing realization of the importance of water management in Jordan from both scientists and politicians during the last decade (Abushandi and Merkel 2011b). The water availability is the most vital factor controlling the economic growth in Jordan as a country based on agriculture and cropping (Figure.C 1). This situation drove the Jordanian government to implement new projects to provide Jordan with fresh water. The proposed Dead Sea - Red Sea canal project is one of these projects to produce hydro-energy, increase the level of Dead Sea to the state of 1960, and produce fresh water. The length of this channel would be about 325 km canal length (Beyth 2007; Hadadin et al 2010). The second proposed project is the exploitation of the Disi aquifer groundwater in the southern desert of Jordan by means of a 2000mm diameter pipeline with a length of 300 km to convey additional water (100106 m3y-1) to Amman (Abu Qdais and Batayneh 2002). In addition, the government tries to enhance the use of treated wastewater for irrigation instead of groundwater. Moreover, water harvesting at the catchment scale is one of the solutions to increase water availability for agriculture (Abdulla et al 2002; Al-Adamat et al 2010; Alkhaddar et al 2005).

112

Figure.C 1 Water use percentage in Jordan (Hadadin et al 2010)

The Wadi Dhuliel arid catchment shares general arid regions characteristics, which can be summarized by: (v) stream flow is characterized by flash flooding during high intensity rain events only, absence of base flow, and low flood frequency components, (vi) (vii) high evaporation rates, spatial rainfall distribution is highly localized (Wheater et al 1991) and the correlation coefficient of storm rainfall decreases more rapidly with distance (Osborn et al 1979), (viii) plant cover and organic matter are sparse (McIntyre and Al-Qurashi 2009; Pilgrim et al 1988). However, studying the complexity of hydrological process in arid region is basically based on understanding rainfall characteristics and watershed properties. The research community is facing challenges of applying rainfall runoff models in arid zone. The main challenge is the lack of monitored data (Pilgrim et al 1988) specially rainfall spatial distribution over the catchment area, because the rainfall is

113

the primary input in any hydrological model. This may force the researcher to minimize the research questions or will increase the model uncertainty. In hydrological modelling, conceptual-lumped rainfall runoff models usually require less input data than distributed models. However, a distributed rainfall runoff model may offer a better approach for flood hydrograph simulation in catchments characterized by the heterogeneity of rainfall distribution (Yu and Jeng 1997). Consequently, spatial rainfall dataset is required for a successful distributed rainfall runoff model analysis. In the present time, the availability of spatially distributed data sets such as Digital Elevation Model (DEM), Landsat 7 Enhanced Thematic Mapper Plus (ETM+), and soil type allow the regionalization of model parameters. Additionally, the open access into rainfall data obtained from satellite-borne derive microwave radiometers is increasing the use of physically-based distributed models. It has been clearly stated the advantage of using Geographic Information System (GIS) in hydrological management by many researchers (Jayakrishnan et al 2005; Martin et al 2005; Reinelt et al 1991). Efforts have also been made on integrating some of hydrological models with GIS environment. Most of these models are physically based distributed, e.g. HEC-HMS, SWAT, TOPMODEL, and WEPP model. This integration allows assessing and predicting the impact of watershed management practices (Arnold et al 1998a; Verma et al 2010; Wheater et al 1999; Zhang et al 2008). On other hand, lumped rainfall runoff models require less input data. Furthermore, the analysis can be performed much faster in Lumped model. The IHACRES rainfall-runoff model (Jakeman and Hornberger 1993) has been successfully applied to several climatic zones.

114

The Hydrologic Engineering Center-the Hydrologic Modeling System (HECHMS) is a physically based distributed model, designed to simulate the rainfall-runoff processes of dendritic watershed systems (Scharffenberg and Fleming 2010). The model was developed by the US Army Corps of Engineers. HEC-HMS model has been widely applied for humid, tropical, sub tropical and arid watersheds to simulate and forecast stream flow. Previous studies on HEC-HMS proofed the ability of HECHMS model to simulate and forecast stream flow based on different datasets and catchment types (Anderson et al 2002; Bournaski et al 2009; Chu and Steinman 2009; Cydzik and Hogue 2009; Knebl et al 2005; Yusop et al 2007). More closely, a study conducted by (Al-Abed et al 2005) on Al-Zarqa Basin using monthly streamflow proved that the HEC-HMS model gave more acceptable results than other models. In contrast, a study by Abushandi and Merkel (2011b) proved that the IHACRES rainfall-runoff model is applicable in the Jordanian arid areas. According to the obtained results, the IHACRES model was able to adequately simulate streamflow in arid catchments when applying the model on a storm event scale. In this context, it is important to note that the results quality depends on the chosen time interval. In this research paper, the IHACRES and HEC-HMS rainfall runoff models were selected for several reasons: (i) model availability and structure, (ii) data availability, and (iii) model applicability in arid catchments. The model IHACRES (Identification of unit Hydrograph And Component flows from Rainfall, Evaporation and Stream flow data) is a simple model, parametrically efficient, and statistically rigorous (Dye and Croke 2003). The IHACRES input requirements are only precipitation and temperature, and streamflow for calibration purpose. The second model, HEC-HMS (Hydrologic Engineering Centers Hydrologic Modeling System),

115

requires inputs for the basin model, meteorologic model, and control specifications but then allows several different techniques to be used to model the rainfall-runoff process. To answer the question of what parameter levels will lead to a desired modelled output, the sensitivity analysis (SA) will be carried out to study the behaviour of modelled streamflow with respect to the change of parameter values. This method is useful for complex hydrological models that involve a large number of parameter (Liu and Sun 2010). In addition, it is particularly important for arid catchment to identify the local controlling parameters. The examination of these two models will based on the application into a single heavy rainstorm event that caused streamflow obtained from the Wadi Dhuliel sub-basin on 29./31.01.2008 by using hourly rainfall, temperature and streamflow data. The key issue of the paper is evaluating whether the distributed HEC-HMS model perform sufficiently better than the lumped IHACRES model. The objectives of this study are to (i) proof the ability of HEC-HMS and IHACRES rainfall-runoff models to simulate stream flow for the Wadi Dhuleil from a single storm event, (ii) parameterize the Wadi Dhuleil using a high resolution dataset including the Global Satellite Mapping of Precipitation (GSMaP_MVK+) and ASTER data, and finally (iii) correlate land activities to water variability through HEC-HMS model.

116

Materials and method

Study area description and data analysis

The Wadi Dhuliel arid catchment is located in the northeast part of Jordan as a major sub-basin of the Al-Zarqa Basin (Figure.C 2). The total area of the Wadi Dhuliel drainage network is around 2687 km2. The size of the area may differ slightly from one author to another basically because of the resolution of the DEM image used to determine the catchment area, the delineation method, and the software used for the catchment delineation. Al-Zarqa, a city of one million people, is located in the outlet section of the Wadi Dhuliel Sub-basin. The climate in the area is arid with an average rainfall of 123mm per annum.

Figure.C 2 Location of the Wadi Dhuleil arid catchment area in Jordan (ASTER, 30m resolution, Map datum: WGS 84)

117

Land use and soil types

The

ecological

infrastructure

such

as

vegetation

cover

type,

soil

characteristics, plant and settlement densities affect the infiltration characteristics and influence the storage coefficient and runoff behaviour. Derived from Enhanced Landsat Thematic Mapper Plus (ETM+) imagery, the land use categories of the Wadi Dhuliel area were classified into six classes (Figure.C 3) ERDAS Imagine (V8.4) software was used to develop a land use supervised classification. This procedure was based on the Maximum Likelihood Classification method with training area based on ground survey and land cover maps (Table.C 1). The ground cover is largely cultivated area in the upstream of the catchment and contributes around 36.2% from the total area. Bare soil and settlements are the main surface cover in the downstream.

Figure.C 3 Land use map of the Wadi Dhuleil (ETM+ 10.04.2005) Table.C 1 Land use features for each sub basin derived from ETM+ 10.04.2005

118

Land use type (%)

Orchard

Seasonal cultivation

Bare soil

High land

Basalt

Settlements

Sub basin No.

1 2 3 4 5 6 7 8 9 Basin total (%)

1 3.2 1.2 0.9 1.1 5.1 0.4 0.5 1.6 1.2

11.8 57.5 44.1 42 43.5 39.1 27.4 36.6 37.5 40.0

0 12.6 27.7 24.2 28.6 18.3 31.9 21.5 17.3 25.4

66.2 14.8 0.1 15.2 0.6 1.0 1 0 0.2 4.1

0 4.9 17.1 7.7 17.2 3.7 31.3 0 10.9 15.0

21.2 7.1 9.7 10 9.0 32.7 8.0 41.4 32.6 14.4

Most of the soil types within the study area is classified as Aridisols and contains high amount of lime (Al-Qudah 2001) low gypsum and basalt in the subsoil and in the parent materials. Soil physical properties and it is relationship to soil moisture have important implications in water flow potential. Spatial soil physical properties data (e.g. sand, silt, and clay %) are obtained from the soil and terrain database, regional project (SOTER) developed by FAO and the International Soils References and Information Center (ISRIC) (Figure.C 4 and Table.C 2). Soil crusting most often occurs due to the high silt content in this type of soils

119

Figure.C 4 Soil classification maps based on ISRIC

120

Table.C 2 Soil Texture and percent composition (sand, silt, and clay) for each sub-basin
Sub basin No. Clay [%] Silt [%] Sand [%] Textural Triangle*

1 2 3 4 5 6 7 8 9 Average

17 21 24.2 21 21.25 23.4 19.8 22 22.7 21.4

42 48 55.7 47 47.25 54.3 49.6 55 55.7 50.5

41 31 20.1 32 31.5 22.3 30.6 23 21.6 28.1

Loam Loam Silt loam Loam Loam Silt loam Loam Silt loam Silt loam Silt loam

*Textural classification in various soil classes (Jackson 1965)

The main wet season is between October and March, while the dry months are between April and September. Topographic data were generated from Advanced Spaceborne Thermal Emission and Reflection Radiometer images (ASTER), and show that the area is characterized by a gently undulating with an elevation varying from 460 m in the southwest to 1400 m in the north (Figure.C 5) with an average slope between 5-30%. The area is considered as the main source of agriculture in Jordan arid regions and groundwater is the main source of irrigation. Irrigation for agriculture in the Wadi Dhuliel taxes existing groundwater resources and strongly distorts hydrological features of this arid catchment.

121

Figure.C 5 The Wadi Dhuleil sub basins altitude (ASTER, 30 m resolution) and GSMaP_MVK points

HEC-HMS model description

The HEC-HMS can be used to simulate a single watershed or a system of multiple hydrologically connected watersheds. The first step in the application of HEC-HMS is to define the basin area and sub basins, a stream network, and diversions, junctions. As any physically based hydrologic model, HEC-HMS simulate most of the key hydrologic processes at watershed scale. The HEC-HMS model requires different datasets including Digital Elevation Model (DEM), weather data, soil type, and land use. A detailed map of land surface elevation was obtained from ASTER with 30 m resolution (Figure.C 5). A comprehensive description of all components in HEC-HMS can be found in the user manual (Scharffenberg and Fleming 2010). The Geospatial Hydrologic Modeling Extension (HEC-GeoHMS) along with ArcHydro extension in ArcView was used to delineate the physical properties from ASTER data and generate a stream network. HEC-GeoHMS was also used to create

122

the input file in the form of sub-catchment boundaries, meteorologic model etc. for the use in HEC-HMS. Figure.C 6 is showing the output of HEC-GeoHMS with the basin being divided into 9 sub-catchments. The Sub-catchments characteristics are summarized in Table.C 3.

Figure.C 6 HEC-GeoHMS generated the Wadi Dhuliel catchment (9 sub-catchments) Table.C 3 Sub-catchments characteristics
Sub-catchment No. 1 2 3 4 5 6 7 8 9 Area [km2] Mean Altitude [m] Total Rainfall [mm]* Number of GSMaP Points

96 508.2 596.1 258.7 575.4 84.01 251.4 189.46 127.83

1100 825 580 735 520 490 663 605 483

22.6 26.8 76.1 17.7 133.7 50 106.2 19.7 21.2

2 3 5 1 6 2 5 1 1

*Rainfall for three days period derived from GSMaP_MVK+ (mm)

123

The HEC-HMS model includes three main components: basin model, meteorologic model, and control specifications. Basin model stores the physical datasets describing the catchment properties. Meteorologic model includes the precipitation; evapotranspiration and snowmelt data. Six different historical and synthetic precipitation methods, two evapotranspiration methods, and one snowmelt method are included. The time span of a simulation is controlled by control specifications including a starting date and time, ending date and time, and computation time step. However, the stream flow simulation requires carful identifications of each model. The loss rate model which has been used in this study is the Soil Conservation Service (SCS) Curve Number (CN) method to compute the volume of streamflow. The SCS-CN method accounts for most of the runoff producing watersheds characteristics such as soil type, land use, hydrologic condition, and antecedent moisture condition (Mishra and Singh 2004) using the following formula: Pe = (P I a ) 2 P Ia + S Equation 1

Where Pe is the accumulated rainfall excess at time t ( Pe = 0 if P < 0.2 S ), P is the accumulated rainfall depth at time t, I a is the initial abstraction, S is the potential maximum retention. The SCS developed an imperial relationship between I a and S as I a = S , where = 0.2 , and then the equation 1 will be:

Pe =

( P 0.2S ) 2 P + 0.8S

Equation 2

Where the parameter S is mapped to the curve number CN as

124

S=

(1000) 10 CN

Equation 3

The model of direct flow which includes the transformation of precipitation into surface runoff was accomplished by SCS Unit Hydrograph. Since the area has no base flow, the model of base flow was not used. Transform method requires a lag time determination as an input. The SCS developed a relationship between the time of concentration ( Tc ) and the lag time ( Tlag ). The time of concentration can be estimated based on subbasin characteristics including topography, length of the reach (Kirpichs formula). Tlag = 0.6Tc
L0.77 Tc = 0.0078 S 0.385

Equation 4 Equation 5

Where, L is the reach length in feet, and S is slope in %.

IHACRES model description

IHACRES is the abbreviation of Identification of unit Hydrograph And Component flows from Rainfall, Evaporation and Streamflow. IHACRES (Jakeman and Hornberger 1993; Jakeman et al 1990) is a hybrid conceptual-metric model, using the simplicity of the metric model to reduce the parameter uncertainty inherent in hydrological models (Croke and Jakeman 2004). The main objective of IHACRES is to characterize catchment-scale hydrological behavior using as few parameters, often about six parameters. The IHACRES model only requires three data sets (rainfall, temperature, and streamflow) per time unit.

125

The classic redesigned IHACRES version(Croke et al 2006; Jakeman and Hornberger 1993) has been used in this study. The original structure of IHACRES includes non-linear and linear modules. The non-linear loss module converts rainfall ( rk ) into an effective rainfall ( uk ) by considering both the infiltration rate and evapotranspiration. In order to obtain the effective rainfall, a catchment wetness index or antecedent precipitation index, representing catchment saturation, is calculated for each time step. Usually, a non-linear loss module within IHACRES includes three parameters: c is the adjustment parameter, f is a temperature modulation parameter and w( const ) is the rate at which catchment wetness declines in the absence of rainfall. The initial stage is to determine the drying rate w , and the catchment moisture index S k at each time step, which is given by:
w = w(const ) e ( 20 t k ) f

Equation 6

where w is the drying rate at each time step, w( const ) is the rate at which catchment wetness declines in the absence of rainfall, tk is the temperature at time step k and f is a temperature modulation parameter (C1), which determines how

w changes with temperature.

Catchment wetness index S k is computed for each time step on the basis of recent rainfall and temperature records. The loss module is used to account for the effect of antecedent weather conditions on the current status ( S k ) of soil moisture and vegetation conditions, and evapotranspiration effects.
S k = c rk + (1 1 ) S k 1 w( k )

Equation 7

126

where c is the adjustment parameter and controls the amount by which S k is increasing by a rainfall event (Post and Jakeman 1999), rk is the rainfall at time step k. Finally the effective rainfall ( rk ) in the model is given by: u k = rk s k if rk s k 0 Since the study area is characterized by ephemeral streams with no runoff if there is no rainfall, computing effective rainfall if rk is >0 for ephemeral streams can make equation (3) true, both physically and dimensionally. In the linear routing module, the effective rainfall is converted into streamflow ( Qk ). The linear model employs discrete-time intervals, transfer function and a representation of the Unit Hydrograph (UH).
Q(qk ) = qQ(qk 1) + qu( k ) Q(sk ) = sQ(sk 1) + su( k )

Equation 8

Equation 9 Equation 10

s where Q q k , Qm are quick and slow streamflow components. Delta in the u( k ) is

the delay between rainfall and streamflow response. The parameters ( q , s ) are the recession rates for quick and slow storage, whereas the parameters ( q , s ) represent the fraction of effective rainfall. The UH of total streamflow is the total of both quick and slow flow UHs.
Rainfall data

The data set of single heavy rainstorm event that caused streamflow in the Wadi Dhuliel catchment was collected from the Surface Water Resources Unit at the Jordan Ministry of Water and Irrigation (JMWI). The only available climatic data from the 127

JMWI are based on a daily time resolution. However, the hourly data sets were obtained directly from the raw chart of Umm-Aljimal meteorological station (Figure.C 7).

Figure.C 7 Rainfall bar chart for a single storm event in mm (Umm-Aljimal meteorological station, JMWI.)

To overcome the problem of spatial rainfall dataset scarcity and ensure the uniformly distribution over the catchment area, re-adjusted satellite derived rainfall dataset known as the Global Satellite Mapping of Precipitation (GSMaP_MVK+) version 4.8.4 was used to determine the rain storm location. It has also been used to determine the ratio of rainfall in each sub-catchment. The GSMaP_MVK+ dataset of a heavy storm event in January 29-31 with 26 grid points covering the basin was used (Figure.C 8). The GSMaP project started in 2002 with support of the Japan Science and Technology Agency (Ushio et al 2009) to produce high resolution rainfall dataset. A frame from 31.95oN - 32.55oN and 36.15oE 36.85oE was extracted from the GSMaP_MVK+ to cover the entire area Wadi Dhuliel catchment with 24 knots and a

128

spatial resolution of 10.8 km. To consider the spatial distribution of rainfall, standard Kriging interpolation (linear variogram) was used to calculate estimates for the representative area of each GSMaP_MVK+ pixel (Figure.C 8). Detailed information about GSMaP_MVK+ dataset processing and application for the Wadi Dhuliel are given by Abushandi and Merkel (2011b). Furthermore, the GSMaP_MVK+ datasets show that the highest magnitude of rainfall was during the first day of the storm (29.01.2008) (Figure.C 9).

Figure.C 8 GSMaP_MVK+ estimates using standard Kriging interpolation for rain storm between 29-31.01.2008, and the average for this storm event respectively

129

Figure.C 9 Average rainfall rates from all GSMaP_MVK+ pixel values for the three days storm (29-31.01.2008)

Streamflow data

Generally, the runoff production in the Wadi Dhuleil arid catchment is totally different from one storm to another. Therefore, the capability of simulating individual storms is important for models to adequately capture hydrologic processes at different conditions. The rate of rainfall was found to be non-linearly related to streamflow on the rising limb of the stream hydrograph. The scarce streamflow gauging is common in the Wadi Dhuleil. Only one streamflow gauge exists in the Wadi Al-Zaatri (subbasin no. 3) with daily dataset covering limited periods. Hourly data can be extracted from the row charts. Due to the limitation of calibration and validation data, hourly streamflow data from Al-Zaatari gauging station was used to calibrate the HEC-HMS and IHACRES models. The objective of the model calibration is to match simulated streamflow volumes with the

130

observed ones. Flash flood hydrograph is generally sharp with timing between 3-24 hours during and after the rain storm (Figure.C 10).

Figure.C 10 Stream flow event in January 30-31, 2008, (Al-Zaatari Gauging Station, Jordan Ministry of Water and Irrigation)

It is recognized that the stream flow volumes in the Wadi Dhuliel have recently very limited magnitudes and dramatically decreased over time. For example, in the years between October 1986 and February 1992, the average flood was 7.6 m3/Sec., while in the years between 2001 and 2008 was 1.2 m3/Sec. This might happen due to urbanization, increasing agricultural activity especially in the upper part of the catchment, and/or drought events. The runoff coefficient in the study area was 2.3% on average (Abushandi and Merkel 2011a). A single rainfall-runoff event dataset in 29-31.01.2008 was used to calibrate and validate HEC-HMS and IHACRES models. Parameter values were tested manually during the calibration processes to reach the best fit between observed and simulated values. This calibration was performed by applying different curve numbers in the HEC-HMS simulated model.

131

The HEC-HMS offers automated and manual calibration. In this study the automated calibration procedure was used. The Nash-Sutcliffe efficiency (Ef) was used to quantify the goodness-of-fit between modelled streamflow to observed records:
Ef =1

(q q ) (q q )
i i i

Equation 11

i is simulated streamflow and q is the mean value where qi is observed streamflow, q of observed streamflow. The relative sensitivity analysis (R) was carried out using the following equation (AlAbed et al 2005):
R = [(( FY 2 FY 1) / Y ) / (( FX 2 FX 1) / X )]

Equation 12

where FY1 is the output result for the original case, FY2 is the output result for the new parameter with specific change, FX1 is the original parameter value, and FX2 is the new parameter value with specific change.

132

Results and Discussion

Based on the hydrologic soil groups (HSGs) and vegetation type, the CN value of each subbasin was determined (Table.C 4)
Table.C 4 Soil Curve Number (CN) method parameters values for each sub basin
Sub No. 1 2 3 4 5 6 7 8 9 basin Curve Number Soil Type Land Use Description*

89 82 87 85 86 87 87 86 88

Loam Loam Silt loam Loam Loam Silt loam Loam Silt loam Silt loam

High land , settlements, and seasonal cultivation Seaosnal cultivation, high land, and bare soil Seasonal cultivation, bars soil, and baslat Sesonal cultivation , bare soil, and high land Sesonal cultivation, bare soil, and basalt Sesonal cultivation, settlement, and bare soil, Baslat, bare soil, and sesonal cultivation Settelments, Sesonal cultivation, and bare soil Sesonal cultivation, Settements and bare soil, basalt

*only main land use cover is presented

The Initial loss, Imperviousness, and the lag time of concentration have also been estimated based on subbasin characteristics including topography, length of the reach (Table.C 5)

133

Table.C 5 Transform model parameter values

Sub basin No. 1 2 3 4 5 6 7 8 9 Initial Loss [%] Imperviousness [%] Slope [%] Lag [h] Time

24.7 43.9 29.9 35.3 32.6 29.9 29.9 32.6 27.3

21.2 12 26.8 17.7 26.2 36.4 41.4 43.5 29.4

25 23 21 18 15 9 17 16 5

3.5 10.7 7.4 8.5 6.8 5 9.3 6.6 8.1

The hydrologic soil groups (HSG) classification reflects soil characteristics and the range of infiltration rates (Table.C 6). Based on this classification, the infiltration rate in the Wadi Dhuliel catchment was assigned to group C.
Table.C 6 Summary of HSG characteristics (McCuen 1997)
Group A B C D Minimum Infiltration Rate [mm/hr] 7.6-11.4 3.8-7.6 1.3-3.8 0-1.3 Surface Runoff Potential Low Moderately low Moderately high High

The shape of the simulated hydrograph is generally following the observed hydrograph in HEC-HMS Model (Figure.C 11), while the application of IHACRES to the same data sets showed some sever weaknesses and overestimation (Figure.C 12). Generally HEC-HMS, with few exceptions, tends to overestimate observed streamflow only slightly. Parameters for best fit of HEC-HMS are listed in Tables.C 4 and 5, while best fit parameters of IHACRES are listed in Table.C 7.

134

Table.C 7 Optimized IHACRES parameter values C 0.00

w( const)
40

F 0.15

q
0.05

q
0.04

s
1

s
0

The relative sensitivity analysis was carried out by adjusting different parameter values in both HEC-HMS and IHACRES models for the subbasin no 3 (where the discharge station is located). After running the models repeatedly, the simulated streamflow results were compared with monitored values at each change of parameters. The most sensitive parameters in HEC-HMS model were the CN with an average of 2.7, while the imperviousness relative sensitivity value was 3.1. The results, however, showed sensitivity to input initial loss parameter (3.3). The rest of parameters in the HEC-HMS model showed a weak effect on the modeled outputs. In contrast, most of the IHACRES model parameters are having a strong influence on the modeled outputs except the adjustment parameter (c). The most sensitive parameter in IHACRES was q (0.5)

Figure.C 11 Wadi Dhuliel catchment observed vs. simulated streamflow hydrographs for the calibration 30-31/01/2008) using HEC-HMS

135

Figure.C 12 Wadi Dhuliel catchment observed vs. simulated streamflow hydrographs for the calibration 30-31/01/2008) using IHACRESl

To estimate the goodness of fit between observed stream flow ( qi ) and i ) was estimated using the Nash-Sutcliffe efficiency ( E f ). modeled stream flow ( q A good performance of HEC-HMS was obtained with E f equal to 0.88, while E f was 0.51 in IHACRES application. This, however, shows a poor performance of the IHACRES model on hourly basis application. The applications of HEC-HMS model is considered to be satisfactory. HEC-HMS and IHACRES are models widely applied for estimating streamflow. The major advantage of employing a GIS based approach in rainfallrunoff modeling is the use of different land use patterns especially in regions with complex mix of land use features and different soil types. While the major advantage of using IHACRES is the minimal input data requirements. HEC-HMS and IHACRES models were applied in this study to calculate the streamflow volume in a single storm of the Wadi Dhuliel arid catchment. GSMaP_MVK+ dataset was used to

136

determine the rain storm location. However, it has also been used to quantify the amount of the rain in each subbasin. Estimated and observed streamflow volumes of a single event was close enough to assume the applicability of the HEC-HMS model approach for the region. However, this was not the case in IHACRES model performance.

Acknowledgements
The authors are thankful to Jordan Ministry of Water and Irrigation, particularly the Surface Water Unit. Sincere thanks go to all members of the GSMaP project for providing the GSMaP dataset.

Reference:
Abdulla FA, Amayreh JA, Hossain AH (2002) Single event watershed model for simulating runoff hydrograph in desert regions. Water Resources Management 16:221-238. Abu Qdais HA, Batayneh F (2002) The role of desalination in bridging the water gap in Jordan. Desalination 150:99-106. Abushandi E, Merkel B (2011a) Application of IHACRES rainfall-runoff model into the Wadi Dhuliel arid catchment (in-press). Journal of Water and Climate Change. Abushandi E, Merkel B (2011b) Rainfall Estimation over the Wadi Dhuliel Arid Catchment, Jordan from GSMaP_MVK+. Hydrol. Earth Syst. Sci. Discuss. 8:1665-1704. Al-Abed N, Abdulla F, Abu Khyarah A (2005) GIS-hydrological models for managing water resources in the Zarqa River basin. Environmental Geology 47:405-411.

137

Al-Adamat R, Diabat A, Shatnawi G (2010) Combining GIS with multicriteria decision making for siting water harvesting ponds in Northern Jordan. Journal of Arid Environments 74:1471-1477. Al-Qudah B (2001) Soils of Jordan. In Zdruli S, Montanarealla L (eds), Soil Resources of Southern and Eastern Mediterranean Countries: CIHEAMIAMB, pp 127-141. Alkhaddar RM, Sheehy WJS, Al-Ansari N (2005) Jordan's water resources - Supply and future demand. Water International 30:294-303. Anderson ML, Chen ZQ, Kavvas ML, Feldman A (2002) Coupling HEC-HMS with atmospheric models for prediction of watershed runoff. Journal of Hydrologic Engineering 7:312-318. Arnold JG, Srinivasan R, Muttiah RS, Williams JR (1998) Large area hydrologic modeling and assessment - Part 1: Model development. Journal of the American Water Resources Association 34:73-89. Beyth M (2007) The Red Sea and the Mediterranean-Dead Sea canal project. Desalination 214:365-371. Bournaski E, Iliev R, Kirilov L (2009) Hec-Hms Modelling of Rainstorm in a Catchment. The Mesta Case Study. Comptes Rendus De L Academie Bulgare Des Sciences 62:1141-1146. Chu XF, Steinman A (2009) Event and Continuous Hydrologic Modeling with HECHMS. Journal of Irrigation and Drainage Engineering-Asce 135:119-124. Croke BFW, Andrews F, Jakeman AJ, Cuddy SM, Luddy A (2006) IHACRES classic plus: A redesign of the IHACRES rainfall-runoff model. Environmental Modelling & Software 21:426-427.

138

Croke BFW, Jakeman AJ (2004) A catchment moisture deficit module for the IHACRES rainfall-runoff model. Environmental Modelling & Software 19:15. Cydzik K, Hogue TS (2009) Modeling Postfire Response and Recovery using the Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS). Journal of the American Water Resources Association 45:702-714. Dye PJ, Croke BFW (2003) Evaluation of streamflow predictions by the IHACRES rainfall-runoff model in two South African catchments. Environmental Modelling & Software 18:705-712. Hadadin N, Qaqish M, Akawwi E, Bdour A (2010) Water shortage in Jordan Sustainable solutions. Desalination 250:197-202 492. Jackson ML (1965) Physical Chemistry and Mineralogy of Soils.1. Soil Materials. American Mineralogist 50:815-&. Jakeman AJ, Hornberger GM (1993) How Much Complexity Is Warranted in a Rainfall-Runoff Model. Water Resources Research 29:2637-2649. Jakeman AJ, Littlewood IG, Whitehead PG (1990) Computation of the Instantaneous Unit-Hydrograph and Identifiable Component Flows with Application to 2 Small Upland Catchments. Journal of Hydrology 117:275-300. Jayakrishnan R, Srinivasan R, Santhi C, Arnold JG (2005) Advances in the application of the SWAT model for water resources management. Hydrological Processes 19:749-762. Knebl MR, Yang ZL, Hutchison K, Maidment DR (2005) Regional scale flood modeling using NEXRAD rainfall, GIS, and HEC-HMS/RAS: a case study for

139

the San Antonio River Basin Summer 2002 storm event. Journal of Environmental Management 75:325-336. Liu Y, Sun F (2010) Sensitivity analysis and automatic calibration of a rainfall-runoff model using multi-objectives. Ecological Informatics 5:304-310. Martin PH, LeBoeuf EJ, Dobbins JP, Daniel EB, Abkowitz MD (2005) Interfacing GIS with water resource models: A state-of-the-art review. Journal of the American Water Resources Association 41:1471-1487. McCuen RH (1997) Hydrologic analysis and design. Prentice Hall,New Jersey McIntyre N, Al-Qurashi A (2009) Performance of ten rainfall-runoff models applied to an arid catchment in Oman. Environmental Modelling & Software 24:726738. Mishra SK, Singh VP (2004) Long-term hydrological simulation based on the Soil Conservation Service curve number. Hydrological Processes 18:1291-1313. Osborn HB, Renard KG, Simanton JR (1979) Dense Networks to Measure Convective Rainfall in the Southwestern United-States. Water Resources Research 15:1701-1711. Pilgrim DH, Chapman TG, Doran DG (1988) Problems of Rainfall-Runoff Modeling in Arid and Semiarid Regions. Hydrological Sciences Journal-Journal Des Sciences Hydrologiques 33:379-400. Post DA, Jakeman AJ (1999) Predicting the daily streamflow of ungauged catchments in SE Australia by regionalising the parameters of a lumped conceptual rainfall-runoff model. Ecological Modelling 123:91-104. Reinelt LE, Velikanje J, Bell EJ (1991) Development and Application of a Geographic Information-System for Wetland Watershed Analysis. Computers Environment and Urban Systems 15:239-251.

140

Scharffenberg W, Fleming M (2010) Hydrologic modeling system HEC-HMS v3.2 users manual. USACE-HEC,Davis, USA Ushio T, Sasashige K, Kubota T, et al (2009) A Kalman Filter Approach to the Global Satellite Mapping of Precipitation (GSMaP) from Combined Passive Microwave and Infrared Radiometric Data. Journal of the Meteorological Society of Japan 87:137-151. Verma AK, Jha MK, Mahana RK (2010) Evaluation of HEC-HMS and WEPP for simulating watershed runoff using remote sensing and geographical information system. Paddy and Water Environment 8:131-144. Wheater HS, Butler AP, Stewart EJ, Hamilton GS (1991) A Multivariate SpatialTemporal Model of Rainfall in Southwest Saudi-Arabia.1. Spatial Rainfall Characteristics and Model Formulation. Journal of Hydrology 125:175-199. Wheater HS, Jolley TJ, Onof C, Mackay N, Chandler RE (1999) Analysis of aggregation and disaggregation effects for grid-based hydrological models and the development of improved precipitation disaggregation procedures for GCMs. Hydrology and Earth System Sciences 3:95-108.
Yu P, Jeng Y (1997) A study on grid based distributed rainfall runoff models. Water

Resources Management 11: 83-99. Yusop Z, Chan CH, Katimon A (2007) Runoff characteristics and application of HEC-HMS for modelling stormflow hydrograph in an oil palm catchment. Water Science and Technology 56:41-48. Zhang XS, Srinivasan R, Debele B, Hao FH (2008) Runoff simulation of the headwaters of the Yellow River using the SWAT model with three snowmelt algorithms. Journal of the American Water Resources Association 44:48-61.

141