Executive Summary

This report comprises the Phase 3 deliverable Hydrology Report of the Jeddah Storm Water Drainage Master Plan (1994). The hydrological component of the master plan requires the analysis of both rainfall and wadi flow data. The analysis of rainfall data is required to update the work done by the Institute of Hydrology in 1983 and 1985 for Kattan-Gibb for the master plan of the southern part of Jeddah. Similarly the flood frequency analysis is required to update work done at the same time.Design rainfalls of different durations, intensities and frequencies are required for the city of Jeddah, along with the likely profiles storms will take. In addition, estimates of flood peaks and hydrographs for different design standards are required for the wadis that drain from the hills to the east of Jeddah towards the sea. The purpose of this hydrological report is to provide these data for the master planning study.

Our recommendations for adoption in the hydrological design of the storm drainage system for Jeddah are summarised in Chapter 6. For rainfall, these include:updated storm depth-duration-frequency curves;an updated value for the standard index rainfall, the one hour five year return period

rainfall depth;new storm profiles based on observed rainfall intensities.

For wadi flows, our recommendations include:an updated relationship for deriving the standard index flood, the five year return period

peak flow;an updated regional flood growth curve;a new relationship for estimating flood volume;a new dimensionless hydrograph for defining flood shape.

We believe these relationships make the best use of the available data, and until such time as more data are available, supersede all previous work and provide the most reliable estimates of hydrological variables, for design purposes, in and around Jeddah.

We also make the following recommendations:the installation of a weather radar post overlooking Jeddah should be seriously considered;a flow gauging station on the Wadi Fatima is installed, for example, at the Jeddah to

Makkah road bridge;at least two other gauging stations are installed on the wadis flowing towards Jeddah;a modelling study is undertaken on the Wadi Fatima to assess the depth, direction, and

volumes of the one hundred year flow across the coastal plain.

1 Introduction

On 11 November 1992 significant flooding in several areas of Jeddah resulted from intense rainfall over parts of the city. In some places streets were under a metre of water, causing extensive damage and disruption to travel and other activities. As a result of this event a report on the extent of the flooding, and recommendations for alleviation of the likely effects of a similar storm, was made by Buro-Happold in 1993. One of the recommendations in their report was that a master plan for the storm drainage of Jeddah should be undertaken. This master planning study was therefore commissioned to establish the capabilities of the existing system and plan a drainage system for the whole of Jeddah city.

This report comprises the Phase 3 deliverable Hydrology Report of the Jeddah Storm Water Drainage Master Plan (1994). The hydrological component of the master plan requires the analysis of both rainfall and wadi flow data. The analysis of rainfall data is required to update the work done by the Institute of Hydrology in 1983 and 1985 for Kattan-Gibb for the master plan of the southern part of Jeddah. Similarly the flood frequency analysis is required to update work done at the same time. The purpose of these updates is to make full use of all available data, including the additional data and significant hydrological events experienced in the area during the last ten years.

Design rainfalls of different durations, intensities and frequencies are required for the city of Jeddah, along with the likely profiles storms will take. In addition, estimates of flood peaks and hydrographs for different design standards are required for the wadis that drain from the hills to the east of Jeddah towards the sea. The purpose of this hydrological report is to provide these data for the master planning study. The rainfall and flood components derived in this stage of the design procedure are probably the most critical parts of the storm drainage design.

The report is split into two principal components. The first is the main report which constitutes this document. The second is the appendices. The purpose of the main report is to present an outline of the work done, results achieved, and conclusions and recommendations in a format understandable to the non-specialist in hydrology. The appendices describes the analysis in more detail. It is assumed that the reader of the appendices will be familiar with basic hydrological concepts such as statistical distributions, regional flood frequency analysis, POT analysis etc.

The Scope and Terms of Reference of the hydrological component of the master plan are set out in Chapter 2 of this main report. Chapter 3 describes the geographical setting of the study and introduces the background to the project. An outline of the methodology adopted and results achieved in the rainfall analysis is described in Chapter 4. A similar outline covering the wadi flood estimation assessment is covered in Chapter 5. The results and conclusions are summarised in Chapter 6, while Chapter 7 makes recommendations concerning further hydrological investigations that should be carried out as well as making some suggestions for the improvement of data collection and quality control.

2 Scope and Terms of Reference

The Terms of Reference for the Phase 3 Hydrology Report are contained in Sections 2.9 and 2.10 of the clients Request for Proposals. These state:

2.9 Collect, collate, and update rainfall data and flooding records.2.10 Analyse all catchment areas and wadis draining into the study area, including those

inside and outside the boundaries of the study area; indicating directions of flow and flood risks for the study area.

As a deliverable the proposals simply state that ‘a hydrology report’ is required (Section 3.4, C).

In our Offer of Services we proposed our approach to satisfying the requirements of the Terms of Reference:

Analysis of rainfall data for storm drainage design was undertaken in 1984 as part of a study by Sir Alexander Gibb and Partners, Hydraulics Research Station, and Institute of Hydrology. This study derived duration-intensity-frequency curves for a small area of Jeddah using three rain gauges with an average of seven years of data. We will extend these curves to include the additional data that should now be available, and any other information available.

Imperial College of Science, Technology and Medicine in London was involved in a hydrological study on the south west coast of Saudi Arabia that ended in 1987. This study included the instrumentation of several major wadis with flow gauging stations and autographic rain gauges. The results from the study of these data have been reported on and it is expected that these results will benefit the interpretation of rainfall statistics obtained for this study.

Design of storm drainage facilities to control wadi flows depend on a reasonable estimate of both the flood peak and the hydrograph shape for different return period events. The nature of these events is highly intermittent, with long periods of inactivity between events. We anticipate assessing the flood risks [to the city of Jeddah] by several different methods to ensure a robust, conservative, but defensible approach to the master planning design requirements.

The results of the hydrological and hydraulic modelling will provide design flood hydrographs for all wadis draining into the study area for return periods of 2, 5, and 10 years, and any other periods that we consider to be necessary for the purposes of the study.

3 Background to the Study

The city of Jeddah is located on the western coast of the Kingdom of Saudi Arabia (21o30’N, 39o10’E), Figure 3.1. The city is the main port on the western seaboard of the kingdom and presently occupies an area of some 1,200km2 (60km N to S, 20km W to E), with a resident population of around 2,000,000. This figure swells to around 4,000,000 during the Haj pilgrimage season when visitors arrive on their way to visit the cities of Mecca and Madinah. The resident population in the early 1970s was estimated as 300,000. The city has therefore experienced enormous growth over the last 20 years and development continues apace.

The climate in Jeddah is warm and humid, although dry. The average temperature (1985-93) is 28oC but the average maximum and minimum temperatures vary between 39oC in August to 17oC in February. Average maximum humidity in all months is close to 100%, but the average minimum is around 10%. Average humidity throughout the year is 59%. Prevailing winds are northerly, although occasional strong winds do come from the south. The average annual rainfall (1985-93) is 47.3mm which occurs over an average of 8 rain days per year, primarily between November and January: during the last nine years one rain-day (1.0mm of rain) has been recorded at the MEPA weather station during June, July, August, and September. Figure 3.2 shews the monthly distribution of rainfall and average number of rain days.

The west coast of Saudi Arabia can be considered as a series of 4 zones running NW to SE, roughly parallel to the coast. The first, and most westerly, zone comprises the coastal plain in which Jeddah is located. This is a depositional coralline aquifer sequence of Tertiary-Quaternary age, and stretches the length of the western seaboard of the kingdom. This plain, known as the Tihamah Plain, varies in width from 10km to 100km, and mainly comprises large, fairly flat expanses of gravels and sands. It is bounded on the eastern side by a series of foothills, which can be considered to comprise the second zone. These hills, formed of Pre-Cambrian crystalline basement, range in elevation from 300m to 1,000m (1,000ft to 3,000ft) near the coast, to 1,500m to 2,000m (5,000ft to 7,000ft) towards the west. It is in this range of mountains, sometimes known as the Asir Mountains, that the majority of the Kingdom’s rainfall occurs (Figure 1.3), with major wadis draining several thousands of square kilometres discharging their flows westwards, onto the coastal plains.

At the western edge of these hills is the spectacular Tihamah scarp, which runs the length of the coast south west from Taif. The scarp rises up to 2,000m (7,000ft) from the wadi floors, and varies in elevation between 2,500m to 3,000m (8,000ft to 10,000ft). North of Taif (130km east of Jeddah) the scarp is less well defined and falls more gradually to the west. The scarp generally defines the divide between flows discharging westwards, towards the sea, and those flowing eastwards towards the interior. However, north of Taif several wadis originate by flowing north or north east before turning west towards the coast. Wadi Fatima, which drains into the southern part of Jeddah falls into this category.

The fourth zone comprises the area to the east of the scarp - a flatter, more arid region which drains eastwards. This extends generally towards the interior of the Kingdom, and elevations generally gradually fall away until the east coast is reached.

The city of Jeddah is backed by a range of relatively low hills to the east that range from 150m to 800m (500ft to 2500ft). The western side of these hills comprise a series of wadis of varying size that drains westwards into the city. To the south of the city, a large wadi system, known as Wadi Fatima, drains an area of some 4600km2 from the escarpment at Taif to the coast. The city boundary extends some way into the mouth of the wadi, and development in this area has already commenced. To the north, the courses of wadi systems are less well defined, this area being more dominated by flat coastal plain.

The rainfall along the west coast of Saudi Arabia is generally subject to two major influences. In winter the supply of precipitation is derived from weather systems originating in the north and west (usually the Mediterranean). In the south, moist air from the south west monsoon system penetrates the area during summer months (Wheater et al, 1989). However, the extreme seasonality displayed in Figure 3.2 (and later in this report in the seasonal analysis of rainfall) demonstrates that these events rarely reach as far north as Jeddah. Local climate is modified by the influence of the Red Sea, the hot interior of Saudi Arabia and the orographic effects of the Asir mountains (see Figure 3.3). Recorded and observed rainfall is generally of short duration events (less than 2 hours) with high intensities, although longer duration events (up to around 18 hours) have been recorded. It therefore appears that rainfall is predominantly convective over Jeddah, although orographically induced, particularly further inland towards the escarpment, where annual rainfall is strongly correlated with elevation.

4 Rainfall4.1 Literature Review

Several studies on the rainfall of Jeddah have previously been undertaken. The purpose of these studies was, like this one, for estimating design rainfall for storm drainage planning. Two studies in particular have analysed some of the available rainfall records. These are ‘Study and Design of Stormwater Drainage for Jeddah Southern Area’ by Kattan-Gibb and Hydraulics Research Station, December 1982, and ‘Jeddah Stormwater - Phase II: Flood Estimates for Jeddah Mountain Region’ prepared for Sir Alexander Gibb and Ptnrs by Institute of Hydrology, January 1985. Both of these studies analysed a portion of the available data. The earlier analysis was based primarily on one daily read raingauge in Jeddah and three recording raingauges located between 70km and 130km to the north of Jeddah. The latter study looked at 12 raingauges around Jeddah, mostly located in the upland region east of the escarpment. The primary purpose of this report was to determine flood runoff from the wadis surrounding Jeddah, including the Wadi Fatima, by both flood frequency and rainfall-runoff analysis. It was therefore more concerned with rainfall in the upland region.

In addition to these studies of rainfall in the Jeddah region, several others have been conducted into the rainfall and runoff regimes of the south western area of Saudi Arabia. In 1987 a major study was completed by Saudi Arabian Dames and Moore entitled ‘Representative Basin Study’. This study took place over a four year period during which five drainage basins , and sub catchments within those basins, were relatively intensely instrumented. This included a dense network of recording raingauges, wadi flow gauges, and meteorological stations. The basins were all located to the south of Jeddah, the furthest north being the Wadi Al Lith basin. Three of the basins drain towards the western coast: the remaining two are located to the east side of the escarpment and drain towards the interior. Much useful information was collected during the course of this study, but its duration was too short to aid much in the estimation of relatively rare hydrological events. Since the study was completed records are no longer kept for the gauging stations.

King AbdulAziz University in Jeddah have recently carried out a two year study entitled ‘Flood Estimation and Impact: south-western region of Saudi Arabia’ sponsored by KACST over the period 1989-90. The area covered includes the south western coast of the country but does not come as far north as Jeddah. The project investigated the determination of peak discharges and flood hydrograph characteristics in the drainage basins of the south-western region using traditional methods of flood frequency and regionalisation. The project also encompassed the computation of water surface profiles for floods of specified frequencies for a selected wadi in the Abha area.

Several papers have been published as a result of these latter two studies, containing guidance for the estimation of design criteria within the areas covered by the studies. Unfortunately, neither study stretches north to include the Jeddah region, where annual rainfall is significantly less than observed further south, and relief is not so exaggerated. These factors have the potential to influence both the design rainfall and runoff hydrology. It cannot be assumed, therefore, that their results are directly applicable to the Jeddah region. Additionally, there are several raingauges in the Jeddah region, and wadi flow

gauging stations to the north of Jeddah, which have never been included in regional studies. These data constitute an important addition to the data set in determining local hydrology.

Various other papers have been published on the hydrology of the south west of Saudi Arabia. Most notable of these is an analysis of rainfall characteristics by Wan (1976). This study concerned rainfall along the escarpment ridge and in the jebel to the south of Jeddah. Other papers concerning rainfall and flood estimation in the area have been produced by Nouh (1987, 1988). The conclusions and recommendations from some of these studies have, however, been questioned (eg Wheater et al, 1989).

The conclusion reached in the literature review was that a separate study needs to be undertaken to update the work done in the previous Gibb reports, rather than merely translating the findings of the studies undertaken in the region to the south of Jeddah. A significant amount of additional rainfall data was thought to be available relative to the previous studies on Jeddah region rainfall. In addition, no other studies have included all the raingauges or gauging stations located in the Jeddah region. The updating of the Gibb studies was also a requirement in the client’s terms of reference.

4.2 Data Availability

There are two primary sources for obtaining hydrological data in the kingdom. These are the Ministry of Agriculture and Water (MAW), Water Resources Development Department, based in Riyadh, and the Meteorology and Environmental Protection Administration (MEPA), Scientific and Technical Documentation Department, based in Jeddah. The former are responsible for operating and holding both rainfall and flow data for an extensive network of sites throughout the kingdom, which are collected on a routine basis from the regional MAW offices. The latter collect rainfall and other climate data (temperatures, average wind, could cover etc.) from 28 Ministry of Defence and Aviation (MODA, of which MEPA is a part) sites (mainly airports) around the kingdom. The raingauges located along the west coast of Saudi Arabia are indicated in Figure 4.1.

The MAW operate both storage and recording raingauges, as well as continuous chart recording flow gauging stations. The Hydrology Section in the Water Resources Development Department of MAW is responsible for collecting all rainfall and flow data from their regional offices, verifying the data, and ensuring that the data are archived properly. It also carries out resources studies, and answers requests to provide information and hydrological data. It was the client for the Dames and Moore Representative Basin Study referred to earlier.

Both rainfall and wadi flow data are held are held on a mainframe computer for the period covering 1960-85. However, since 1985 no data have been entered onto computer, which is said to be broken, and more recent data exist only in written form.

A substantial volume of hydrological data has been published. Two volumes, in particular, contain much useful information. The first of these is ‘Monthly Hydrological Data for 1963-80, Rainfall Intensity No 98 Vol 4”, published by the Ministry of Agriculture and Water, Department of Water Resources, Hydrology Division. This contains rainfall intensities for all recording raingauges in the kingdom, for all recorded storms, over the

period 1963-80. The second contains runoff data and is covered in Chapter 5.

The published rainfall data are in form of rainfall depth (mm) in the most intense 10, 20, and 30 minutes, 1, 2, 3, 6, and 12 hours, as well as the storm total. There are a total of 58 of these recording raingauges listed for the south west coast of the kingdom (see Figure 4.1). We were able to obtain data for many of these stations for 1981-85 as listings from the mainframe computer in the same format as that published. However, data for the more recent period was more elusive. We were eventually provided with hand written reduced data for a selection of raingauges, particularly in the region around Jeddah. These data were also in the form of rainfall depth in the most intense periods of specified duration.

It was discovered at an early stage in the project that the Ministry (MAW) have operated a recording raingauge at their offices in Jeddah since October 1987, although this station is not listed anywhere. Daily rainfall totals for this gauge were obtained from the Jeddah office for the period October 1987-93, but the original charts recording rainfall events are sent to MAW in Riyadh. Information concerning storm profiles was required for this study, but was not available in the published volumes which list maximum intensities only. Copies of these charts were therefore eagerly sought from MAW, Riyadh. Initially, knowledge of this raingauge was denied, but after several months we were provided with sight of some of the charts. From these we obtained some storm profiles and maximum five minute intensities for three large events, although most of the annual maximum events were missing.

At around the same time, we learned that the King AbdulAziz University in Jeddah operate a recording raingauge. This site is located around 5km from the MAW regional office in Jeddah on the south side of the city. The only other raingauge known to exist in the city is located at the airport to the north. Data from this gauge would therefore provide invaluable information on the spatial variability of rainfall in Jeddah as well as further information on storm profiles. Unfortunately these data proved particularly difficult to obtain. When we did receive them, at a very late stage in the analysis, it transpired that they were daily totals only, for the period 1987-93. Nevertheless, they were able to provide some information on spatial rainfall variability as well as some additional, independent, 24 hour annual maximum events.

Twelve of the 28 climate stations operated by MEPA are located in the vicinity of the west coast of the kingdom. We were originally led to believe that several of the raingauges at these stations, including the Jeddah airport raingauge, were of the recording type. We were therefore hopeful that we would be able to obtain these data, which would not only increase our understanding of short duration rainfall, but also provide us with invaluable information on storm profiles. We therefore identified the dates of all significant storms (greater than 5mm in a day, or the annual maximum if this was less than 5mm) for each of the 12 raingauges for their period of record, and requested copies of the rainfall charts for these events, where available, and hourly logs where charts were not available.

We were eventually given access to the archives but, unfortunately, it transpired that none of the MEPA raingauges have recording facilities. Instead, rainfall totals are logged, almost incidentally, at varying recording intervals through the record. In the early parts of records, during the 1960s, totals were recorded six hourly, although sometimes notes on

the logs indicated that the rain fell in a shorter duration, the start and finish times of rainfall being recorded. After 1971 the format of the recording logs at the Jeddah station, which has operated since 1960, changed, and rainfall depths were noted hourly. Records were available up to 1993, although events after 1985 had to be transcribed rather than photocopied. These data therefore constituted the most important component in the rainfall analysis of our study. This was also the best information we were able to obtain on storm profiles in the Jeddah region.

A small amount of rainfall data for five raingauges in the Wadi Yiba basin, south of Jeddah, were available to us from an MSc thesis (Brown, 1987). These comprised 10 minute rainfall depths through time for 13 events over the period 1984-85. These, together with the hourly data from the Jeddah raingauge, constituted the basis of the storm profile analysis.

Visits to Jeddah uncovered the existence of some anecdotal accounts of past floods in city. These mainly came from the storm sewer maintenance section of the Municipality of Jeddah. There was also an indication that there may be photographic evidence of a historical flood in 1948 in the Museum of the Municipality of Jeddah. The naval base, at the southern end of the city, is also thought to have had experience of a number of historical floods due to flows from the Wadi Fatima. We have used the reports from the sewer maintenance section to identify areas of flooding. However, we were unable to gain access or to establish contact with the relevant personnel in order to substantiate the other reports.

A detailed list of the rainfall data requested, collected, and used in this study is provided in Appendix A.

4.3 Methodology

There were two principal objectives required of the rainfall analysis in the hydrological study. These were:

Establish rainfall depth-duration-frequency curves applicable to Jeddah for the purposes of storm sewer design.

Recommend storm profiles for distributing the rainfall derived in 1.

Initially it was hoped that regional rainfall relationships could be described for the west coast of the kingdom. However, severe delays and difficulties experienced in obtaining the relevant data, meant that the study had to be trimmed down to the barest minimum of defining a region around Jeddah, in order to achieve the specific project objectives.

The approach adopted was to identify a regional zone of statistically homogeneous rainfall and perform a station year analysis on the raingauges in this zone. This method assumes that stations within a homogeneous region are affected by the same climate and meteorology, and therefore experience a similar pattern of rainfall, although not necessarily the same quantity of rainfall. This approach was necessary because, although there are 30 years of daily rainfall data at the MEPA raingauge in Jeddah, there is little

information on short duration rainfall.

A detailed description of the methodology employed to achieve these objectives may be found in Appendix A, but the following summarises the main steps involved:

collect, collate, and prepare data;exploratory analysis of data;determine homogeneous rainfall regions;perform station year analysis to derive regional depth-frequency-duration relationships;carry out peaks over threshold (POT) seasonal analysis for comparison with station year

analysis;analyse and present results.

The methodology for analysing storm profiles is much simpler due to the paucity of data, which, as described above, were limited to two sources. In this case the following approach was adopted:

collect and collate all available data for storm profiles;group storms into ranges of total rainfall depths and storm durations;define average storm profile for range of storm durations and/or storm depths;tabulate results for use in design.

4.4 Results

This section summarises the main findings of the rainfall analysis of interest to the master planning project. A more detailed description for the interested reader may be found in Appendix A.

4.4.1 General description of rainfall

The principal rainfall events for the Jeddah MEPA raingauge (41024) are presented in Figure 4.2 for the period 1960 to 1993. The dates of events with more than 50mm of rain are indicated on the plot. Several observations can be made from the figure. It can be readily seen that relatively large storm events are not infrequent occurrences in the history of this raingauge. The occurrence of the storm of 2 November 1992 ranks 9th in the 34 year record. However, it is the largest since 1979, and, therefore, in the last 15 years. It follows from this that the recent period has been relatively dry compared to the previous 20 years. In particular, the 1970s experienced an especially wet spell with large (greater than 50mm) events occurring almost every other year. The Buro-Happold report (1993) contains a similar plot although some events were mistakenly indicated. However, a depth of 140mm in 24 hours is indicated on their figure as being recorded some time in the period 1941-46. We have not been able to verify this event and have therefore considered the rainfall distribution both with and without it.

All recorded rainfall events for the three identified Jeddah raingauges for the common period of record are presented in Table 4.1. Annual maximum events for each raingauge

are emboldened. 41024 refers to the MEPA raingauge located in the airport at the northern end of the city. J254 is the MAW reference for the raingauge located in their regional offices towards the south eastern edge of the city. The King AbdulAziz University raingauge (KAU1) is located approximately 5km to the north of J254. The table displays some extraordinary characteristics, if the data are to be believed. It should be noted that all three raingauges are operated by different bodies. If taken at face value, the table indicates an extreme variability in areal rainfall across the city. Further discussion on the contents of this table may be found in Section 4.5.

4.4.2 Rainfall depth-duration-frequency relationships

Of all the raingauges located in the south west of Saudi Arabia, 58 MAW recording raingauges and 3 MEPA raingauges along the coast were selected for analysis. These are shewn in Figure 4.1.

The first task to undertake in analysing regional rainfall characteristics is to define a ‘region’. A rainfall region is an area where characteristics of the rainfall are homogeneous. This requires a knowledge of what aspects of the rainfall are important. For example, the average annual rainfall might be a measure used to ensure that similar rainfall characteristics are experienced within a region. This definition, however, tends to be of limited use because average rainfall can be highly variable over short distances, while the meteorological and climatic conditions giving rise to rainfall are similar over a much larger area: all that changes between many sites is that rainfall is heavier when it does occur. In this study, it is not the depth of rainfall that is particularly of interest, but the relationship of frequency of events to depth of rainfall.

In order to define homogeneous rainfall regions, therefore, we used a ratio relating the depths of rainfall experienced in hour at two different frequencies (once every five and once every ten years) for each site. This gave some indication of the variability of rainfall at each gauge. We then plotted this ratio on a map of the coast to determine areas with common rainfall variability. The homogeneous region we defined using this approach included the whole of the coastal strip of Saudi Arabia from just north of Jeddah to Jizan below an elevation of 100m above sea level, and included 8 raingauges. This area is defined on Figure 4.3.

The average rainfall recorded at each of these raingauges varies, as expected, from site to site within our homogeneous zone. In order to compare the frequency-depth relationship of rainfall at each raingauge, therefore, a means of taking out this background rainfall is required. The usual method is to divide the rainfall depths at each raingauge by, for example, the average rainfall recorded at the gauge. This is called ‘standardising’. In keeping with previous workers in the field, we have used the rainfall depth from a one hour storm which occurs, on average, once every five years for standardising. The reason for using this standard rather than average annual rainfall, is that has a more precise statistical meaning, and is therefore more useful when comparing the frequency relationships of different raingauges. It should also affected less by the number of years of available data than the average rainfall. We have referred to this standard as the ‘1H-M5’

(called the ‘1 hour M5’) rainfall, which is the terminology used in the UK Flood Studies Report.

We estimated this 1H-M5 rainfall for each raingauge from the largest depths of rain that occurred during one hour in each year from a fitted statistical distribution. These largest rainfall depths are termed the ‘annual maxima series’. The 1H-M5 rainfall thus derived was then used to standardise all the annual maximum rainfall depths at each duration for each raingauge in the homogeneous zone. The dates of all the annual maximum depths at each duration, and at each raingauge, were checked and any events that occurred on the same day were removed from the data set. This was done to ensure that the same rainfall event was not counted twice, so ensuring the statistical independence of events.

The standardised data from all the raingauges were then pooled and treated as if they were from the same raingauge. Various types of statistical frequency distributions were fitted to the data for each duration, one of which was finally selected as being the most appropriate. We were able to define formal relationships between the family of distributions for each duration, and these were used to define our final depth-duration-frequency curves. The curves are presented in Figure 4.4.

It should be noted that the rainfall depths along the y-axis of the graph are still in standardised units (‘growth factors’). In order to use them, therefore, a knowledge of the local 1H-M5 rainfall is required to de-standardise them. In their present state, being dimensionless, they may be applied to any location within the homogeneous zone.

The local estimate of the 1H-M5 rainfall depth required for Jeddah is derived from the MEPA raingauge located at the airport. It is 41.6mm. This value is derived after making an adjustment which incorporates the differences between the data available at one hour duration and that available at a 24 hour duration (Figure 4.4). The nature of this adjustment is described in more detail in Appendix A.

The de-standardised storm depths for various durations and return periods for use in Jeddah are presented in Table 4.2. The growth factors for each duration and return period are also listed, as are the parameters of the Extreme Value Type I distribution (EVI) used to define the curves. The equations relevant to the EVI distribution, and equations relating distribution parameters to duration, are included in Appendix A. They allow the relevant depth of rainfall for any duration and any return period to be calculated directly. It should be noted that these equations should not be used for deriving depths of rain outside the range of data from which they were derived.

4.4.3 Storm Profiles

The depth of rain expected to fall, or to be exceeded, with a stated frequency, or rarity, has been defined in the previous section. However, nothing has been said about the distribution of this rainfall, either in time or in space. For design purposes it is necessary to say something about both of these factors.

Historically, design storm profiles through time have been based on a symmetrical profile, where the maximum intensity of the storm peaks midway through the storm’s duration.

The reason for adopting this profile is simply that there has been no data on which to base any other design criteria, and it constitutes a moderately severe storm, relative to a constant intensity profile.

We investigated the shape of storm profiles for 34 raingauge events with depths of at least 10mm. 13 of these events were from the MEPA raingauge at Jeddah, where at storms had to be at least two hours in duration before they provided any information on profiles, since rainfall at this gauge is only recorded on an hourly basis.

We were unable to find any regular patterns looking at events greater than 10mm. However, when considering events greater than 20mm (roughly equivalent to a one hour storm with a two year return period). When split into durations of events lasting up to three hours, and those lasting longer than three hours, it was possible to see a pattern emerging. Splitting the events in this way gave 14 events in the one to three hour group (four of which were derived from the MEPA gauge) and five in the three to nine hours group (one of which was from the MEPA gauge).

In the one to three hour group most events (10, three from the MEPA gauge) started with high intensity rainfall, which either continued until most of the total depth had fallen, or slackened of after around half of the total depth had fallen. However some storms (four, one of which was from the MEPA gauge) started with a low intensity rainfall, which lasted for in excess of fifty per cent of the storm’s duration, followed by a burst of heavy rain at the tail end of the storm in which the majority of the rain fell. We therefore treated these storms as separate groups to derive the mean profile for each. These profiles are shewn in Figures 4. They are referred to as positively skewed storms (front loaded) and negatively skewed storms (back loaded), respectively.

On the observed basis that ten out of 14 events were positively skewed, and four were negatively skewed, we can tentatively say that there is a 2/3 chance of a storm being positively skewed, and a 1/3 chance of it being negatively skewed. The effect of both types of storm would need to be investigated on any design system. However, taking this a step further we have defined a set of combinations of positively and negatively skewed storm events that provide an equivalent return period of failure for a design system, on the basis that negatively skewed storms (back loaded) are more arduous for a system than positively skewed storms. Appendix A contains more details of this joint probability approach.

Of the storms with durations in excess of three hours (eight in total, four of which were from the MEPA gauge) seven shewed a positive skew, although not as severe as observed in the shorter duration events, and one (a Wadi Ghat event) exhibited a negative skew. We therefore treated this latter separately, excluding it from the average profile derived from the other seven storms, Figure 4.8.

The profiles defined as a result of these analyses are summarised in Table 4.3 in terms of percentage of total storm depth to have fallen in a given percentage of the total storm duration.

4.5 Discussion

4.5.1 General Rainfall Characteristics

Table 4.1 presented the rainfall events recorded at the three raingauges in Jeddah for the period 1987-93. It was noted in Section 4.3 that extreme variability in recorded rainfall, both in terms of dates and depths, is displayed. The question must therefore be asked whether such variability is realistic, or whether the data have been wrongly recorded or transcribed.

In most cases the annual maximum events at each gauge occur on different dates (see Table 4.1). Sometimes large rainfall events are recorded on consecutive days at different raingauges. For example, on 10/12/89 27.8mm was recorded at KAU1, where nothing is noted at the Ministry raingauge J254 on that day; on 11/12/89, the following day, nothing was recorded at the university, but 31.6mm was recorded at J254. However, the lack of consistency in this pattern (e.g. on 7/12/88 17.2mm was recorded at J254 with nothing at the university gauge, but on the following day, 8/12/88, 2.8mm was recorded at the university with nothing at J254), precludes the explanation for this lying in different definitions of a rainfall day. Several examples of this incompatibility of records can be seen throughout the table, although on other occasions large events do coincide, e.g. 25/4/90, and, interestingly, 2/11/92.

A further example of this variability, lending support to that displayed in Table 4.1, can be demonstrated from an event recorded in Wadi Ghat in the Yiba basin, in the mountains to the south of Jeddah. This catchment was established as part of the Dames and Moore Representative Basin Study. It was gauged at its outlet, with a drainage area of 597km2. Rainfall over the catchment was described by five autographic recording raingauges, with an average spacing between adjacent raingauges of 13-14km. On 1/5/85 a rainfall event occurred which gave rise to a flood with a peak flow of 50m3s-1. At one raingauge at the edge of the catchment a depth of 10mm was recorded over a period of 3 hours. At two of the adjacent raingauges total depths of 3.2mm and 3.4mm, respectively, were recorded. The two remaining raingauges recorded no rainfall. In order to account for the volume of water recorded in the flood, 80% of the recorded rainfall would have had to runoff (Brown, 1987). Usual percentages of runoff from rainfall in arid regions lie in the range of 10%. Clearly, the main part of the storm fell between raingauges, and thus failed to be recorded.

Other workers in arid regions describe convective rain cells as being very localised, often spreading over only a few square kilometres (Fogel and Duckstein, 1970). In other arid countries, convective storm cells have been described as typically being elliptical, commonly 10km by 5km in size, and lasting around one hour in duration (Simon and Last, 1985). These storms tend to be self-destructive and give rise to short, intense bursts of rain. Wheater et al (1991), looking at hourly rainfall in the Representative Basins in the wadis south of Jeddah found that raingauges even only 8km part rarely recorded rainfall at the same time. They found that, typically, only two raingauges in the five basins would record significant rainfall on the same day. However, they describe another category of storm which shews somewhat more widespread occurrence, in which individual gauges typically record one or two hours of rainfall but with an overall pattern in which rainfall appears to be generated over a group of gauges and to decay and regenerate at a different

location. No evidence of storm movement could be detected from the network.

Evidence of this type of rainfall is displayed in Table 4.1. For example, the event of 2/11/92 occurred over a period of nine hours at the MEPA raingauge (total depth 51.5mm) whereas the rainfall at the Ministry raingauge (27.8mm) fell in 10 minutes, the bulk of occurring in five minutes. In the light of the available evidence, therefore, it is not inconceivable that events and dates recorded in Table 4.1 are, indeed, accurate.

It is very important to understand the nature of these rainfall events, particularly in designing systems to cope with the runoff that they generate. The occurrence of high intensity events over small areas leads to prospect of invalidating the concept of areal reduction factors for these events. In fact, there are grounds for saying that a factor should be introduced to increase the rainfall depths for storms of given durations and return periods. This is on the basis that, because events are so localised, the chances of a raingauge recording an event over an area the size of Jeddah are substantially less than 100%. The frequency-depth relationship observed at a raingauge is therefore likely to underestimate that experienced over the larger area. Past workers have failed to recognise this variability. For example, the Gibb Report (1982) states that an error was found in the Ministry’s raingauge in Jeddah on 3/11/72 on the basis that, because a well-documented 83mm was recorded at the airport raingauge and only 15mm was recorded at the Ministry gauge, the latter must be wrong.

The overriding observation that comes out of this discussion is that the observed rainfall regime over Jeddah is woefully inadequate for the purposes of storm description, both temporally (in time) and spatially (in space). Questions concerning storm durations, intensities, shapes, profiles, areal extent, and frequencies, are unknown. It is unreasonable to expect a single, hourly-read raingauge to provide the necessary information. The station-year analysis, adopted here, is an attempt to address questions relating to the frequency and duration of events, but the ability of raingauges several hundreds of kilometres away to achieve this, despite being in the same defined homogeneous zone, must be questionable.

Given the importance of the city, its increasing population, and its expanding area, a more deterministic means of recording rainfall over the city should be investigated. The most obvious solution to this is some form of weather radar, where all of the above listed unknowns are measured. It is apparent that such a system would record no rainfall over Jeddah for most of the year. Nevertheless, if it was strategically located, it could serve a much wider area, including, for example, Mecca, and possibly also Taif.

4.5.2 Rainfall depth-duration-frequency relationships

The results derived in our study have been based on an objective analysis of the available data. We have tried to limit the assumptions made in order to identify the characteristics displayed by the data. Inevitably assumptions, simplifications, and generalisations do have to be made, and our study is no exception. Nevertheless, there is good evidence to support the main building blocks of our analysis, which lend a degree of confidence to the results.

Comparisons must be made with results and recommendations made in previous studies,

and reasons for any differences must be explained in order for the later study to gain credibility. We have therefore compared out results with those from the two previous Gibb studies (the hydrological analysis, in both cases, being performed by the Institute of Hydrology).

Table 4.4 compares the results from out study (WSA), with those of the two Gibb studies, for various durations and return periods. The depth-duration-frequency relationships from the three studies four one hour and 24 hour durations are presented in Figure 4.9 (the 1982 Gibb study did not provide 24 hour duration totals, and so these are not present on the figure).

For frequent events (two year return period) IH 1985 figures agree remarkably well with ours for all durations. Those from the IH 1982 study are somewhat higher, although still quite similar. As return period increases, events become rarer, and rainfall depths increase. However, our increase at a substantially faster rate than those of IH (1985). The 1982 IH curve for one hour storms is steeper than both ours and the IH (85) curves for return periods up to 10 years. It is interesting to note that the IH (82) rainfall depth for a one hour storm with a ten year return period (65mm) is almost as large as that for the IH (85) 24 hour storm with a ten year return period (67mm).

The reason for the differences in slope can be found in the original data on which the studies were based. The IH (82) study used data from three recording raingauges to the north of Jeddah. The curves which were derived from these data were less steep for a one day duration than the curve indicated for the MEPA Jeddah gauge for which they also had data. The curves from the three then adjusted, by means of comparing ratios, to have a similar gradient to those of the Jeddah raingauge. However, homogeneity between the sets of data used was not demonstrated, and the adjustment was convoluted, adjusting ratios of ratios using more ratios. It is difficult, therefore, to be confident that results so derived actually represent reality.

The raingauges used to derive the IH (85) curves were located around the Jeddah region, mainly in the upland area between the coast and the escarpment. No investigation was undertaken into the suitability of these raingauges to describe the rainfall in the vicinity of Jeddah. Homogeneity, and independence, within a rainfall region were recognised as the underlying assumptions in a station year analysis, but in the time available they were not able to carry out statistical tests to confirm these. Nevertheless, they reviewed the data to confirm that, by and large, annual maximum events were rarely recorded from the same storm at different gauges. They also concluded that, since only the ratios of rainfall at various return periods to the standard rainfall depth (1H-M5) needed to be consistent between gauges, “the region covered by these raingauges was homogeneous and rainfall events sufficiently independent to allow the station year approach to be used.”

However, in our study we used an index of this relationship between ratios to define a homogeneous region. We concluded that raingauges in the upland region display consistently different ratios with the standard rainfall depth than those on the coastal plain. In short, rainfall along the coastal, while being less than that observed in the uplands, is more variable. The gauges used in the IH (85) study were therefore not included in our definition of a homogeneous rainfall region around Jeddah because their rainfall

characteristics were not as variable as those along the coast, including Jeddah. The outcome of this is seen in the regional rainfall curves in Figure 4.9, where the IH curves are less steep than those derived in our study.

In the analysis of west coast Saudi Arabian rainfall carried out by Wheater et al (1989), rainfall variability was found to generally decrease with increasing elevation: the steepest curves from stations located on the west side of the escarpment were those from stations whose elevations of ranged from 10m to 150m above sea level. This generally supports our observations, confirming the variability of rainfall on the coastal plain, relative to that in the uplands.

It should be mentioned that the primary purpose of the IH (85) study was to provide rainfall for hydrograph modelling for the wadis that drain into the south of Jeddah, including Wadi Fatima, and not for storm sewer design in Jeddah. The raingauges used in their study would therefore be more representative of the rainfall regime over the Wadi Fatima catchment as a whole. However, their ability to describe rainfall over the local wadis, where elevations are much lower than in the uplands proper, must be uncertain.

The estimation of the standard 1H-M5 rainfall for Jeddah produces a figure of 41.6mm. This is larger than that used by IH (36.4mm) derived from the upland gauges, but falls within the range observed by Wheater et al (1989) of 19.3mm to 47.5mm.

4.5.3 Storm Profiles

Little appears to have been published on storm profiles in the Arabian Peninsular - we were unable to trace anything through literature searches. However, it is generally observed that convective storms tend to be short duration, frequently with a high intensity burst at the beginning. The adoption of UK design standard storms with symmetrical shape does not, therefore, represent conditions encountered, in this part of Saudi Arabia.

The profiles derived in this study tend to confirm the observations of ‘front-loaded’ storms, although the frequency of ‘back-loaded’ storms in the analysis was surprising. The behaviour of the storm water drainage system is modelled on computer by a software package called SPIDA in this master plan. In SPIDA model runs of the existing storm water system, the difference in effect of these storm profiles, for the same depth of rain, was significant. The back-loaded storms were more severe on the system than front-loaded ones.

For design purposes, the system should be tested with both profiles, since, for example, a two year return period storm of one hour duration could take either profile. The chances of a storm taking either profile do not appear to be even. For this reason, it is not possible to state what, for example, a two year, one hour storm looks like. However, the return period of failure of a storm water system, which is the important factor rather than the return period of a storm, can be estimated by comparing the effects of positively and negatively skewed storms from different return periods on the system. As previously mentioned, Appendix A contains pairs of return periods for positively and negatively skewed storms which, if a design can adequately handle both, will provide the return period of failure of the system.

One question which this analysis raises, and which cannot, at present, be answered, is whether the back-loaded, or negatively skewed, profiles recorded actually represent storm shape, or whether they merely reflect sampling inadequacies. For example, it is entirely conceivable that the low rainfall depth at the beginning of a negatively skewed profile is simply derived from the fringes of a nearby convective cell, whose centre lies some distance away from the raingauge. The high intensity part of the profile could be derived from an entirely different convective cell which happens to develop nearer to the raingauge. If a second cell developed further away from the raingauge, its rainfall would unlikely be recorded, and so a low total rainfall would be observed for the event, which would then be discarded from the analysis. If this is the pattern which gives rise to the observed profiles then separate storms are being treated as one in the above analysis. Unfortunately, sufficient data to discern these effects are not presently available, and so the analysis of observed data must be retained.

In order to simplify the design process, it would be sensible to first design systems using the more common positively skewed storm. Nevertheless, a design should still be tested against a negatively skewed storm of the same depth for the reasons discussed above.

5 Wadi Flows5.1 Literature Review

Much of the discussion in Section 4.1 concerning studies focusing on rainfall analysis applies to wadi flows. Most cover both aspects in more or less detail.

The Gibb report (1985) estimated wadi flows for the southern group of wadis flowing west towards Jeddah. This comprised those contributing to flows into the southern storm water channel, and Wadi Fatima. In estimating these floods, two methods, flood frequency and rainfall-runoff, were used to derive results. These results were then compared and averaged to produce their recommended design flows. The Gibb (1982) study did not estimate wadi floods.

The King AbdulAziz University study focused primarily on wadi floods. The two methods mentioned above were investigated, along with a third method involving hydraulic routing.

The Dames and Moore (1988) study also used both rainfall-runoff and flood frequency analysis. In this case, however, there were insufficient data to carry out a flood frequency analysis on gauged wadi flows. Runoff data were, therefore, generated using a calibrated rainfall-runoff model with a long time series of rainfall which was, itself, generated using a stochastic model of rainfall. This model was developed specifically for the project by Imperial College in London, UK. A description of the model may be found in the published literature (Wheater et al, 1991a, 1991b).

Other workers involved in the design of the storm water system for Jeddah have carried out some flood flow analysis from the surrounding wadis (e.g. Watson Saudi Arabia, 1974). These studies have been based on rainfall-runoff analysis, using storms of specified frequency to predict flood hydrographs and peak flows for specific return periods.

Several other studies have focused on the routing of floods down wadis on the Arabian Peninsular and the effects of transmission losses on both hydrograph shape and peak flow (e.g. Dames and Moore, 1988; Clark and Davies, 1988).

5.2 Data Availability

The primary source of data for any study of flood flows is the stream gauging network, which provides information on gauged, or recorded, flows through time. Wadi flow gauging stations in the kingdom are operated by the Ministry of Agriculture and Water. The Ministry operates an extensive network of gauging stations across the kingdom, many of which are located in the south-west. Figure 5.1 shews the locations of wadis with data available to this study. It should be noted that none of the wadis draining towards Jeddah are gauged, including the 4600km2 Wadi Fatima catchment.

In 1983 the Hydrology Division, Department of Water Resources Development of the Ministry of Agriculture and Water published a summary of all the gauged daily flows for all the gauging stations in the country, entitled ‘Monthly Hydrological Data, Runoff for 1966-80, No 98, Vol 2’. This publication is the only definitive list of gauging stations we

have been able to find. It has not been updated since it was first published.

For this study peak flow instantaneous discharges are required. Daily flow data cannot be used to study floods in arid zones because the floods rise and fall so rapidly that any averaging out over a period of, say, a day, will grossly underestimate the peak. Unfortunately, peak instantaneous flows are not published in the volume referred to above. We were informed by the Ministry, however, that they are kept on the same computer as the rainfall. The status of these data is therefore the same as that of the rainfall: flows up to 1985 are computerised, but since that time no data have been input, although they should be held in archives.

We were unable to obtain a copy of these peak flows from the Ministry until a late stage in the project. We were, however, fortunate in being able to obtain copies of all recorded annual maximum instantaneous flows, for a large number of gauging stations along the west coast of the kingdom, from the Institute of Hydrology in the UK.

We also requested 10 minute hydrographs for annual maximum events during the period 1975-93 for four specific gauging stations located around Jeddah in order to perform a hydrograph shape analysis. Unfortunately, these data were not made available to us.

In addition to the above, we requested copies of rating curves, with any spot gaugings, for Wadis Khulais and Lith. Wadi Khulais is the closest gauged catchment to Jeddah (located to the north), and Wadi Lith (to the south of Jeddah) represents one of the more reliably rated catchments since it was intensively studied in the Representative Basins Study.

The Ministry were reluctant to give us these data since, it became apparent, none of the gauging stations along the west coast of the kingdom have been surveyed, and hence rated, since 1987. It also became apparent that none of the stations set up as part of the Representative Basin Study by Dames and Moore were continued beyond 1987. These two facts represent a major deficiency in the gauging network of the west coast, and therefore of the kingdom as a whole. If reliable data are not collected, reliable results cannot be produced.

The Wadi Fatima has been dammed in its upper reaches by a 600m wide, 15m high concrete gravity dam for the purpose of ground water recharge. The dam, at Abu Hasani, was completed in the mid 1980s, since which time it has been operational. We therefore requested maximum water levels for all events that have occurred since the dam was built. In addition, we requested the depth-volume relationship for the reservoir in order to convert water levels into flood volumes.

Unfortunately, water levels were only available for two events, with the additional comment that in two further years no flood was recorded. The data received were of poor quality and difficult to interpret in that it appeared that various gates in the dam had been operated during the course of the events. Difficulties with interpreting the data were enhanced when it became apparent that the depth-storage relationship was carried out to a datum other than that used by the gauge board up the side of the dam, and that the relationship only extended for the first six metres of water. Thereafter, the reservoir was unsurveyed, and its storage volume was unknown. At least one of the events appeared to fill the reservoir to a depth of 8 metres.

In summary, then, the following points can be made:

there are no flow data for any of the wadis the flow towards Jeddah;data quality for all gauging stations since 1987 must be treated as suspect;no hydrograph data were available for shape analysis;only peak flow data up to 1985 are readily available on computer;data for the Abu Hasani dam on the Wadi Fatima are inadequate for use.

5.3 Methodology

There were three principal objectives that this hydrological study concerning wadi flows aimed to achieve. These were:

estimate peak flows for various return periods for all ungauged wadis that drain in to Jeddah, including the Wadi Fatima;

estimate design hydrographs associated with these peak flows;provide estimates of runoff volumes associated with peak flows.

As none of the wadis draining towards Jeddah are gauged, none of these objectives could be achieved directly using local data. Some other means of relating the hydrological behaviour observed at other gauged wadis in the region to the local ungauged sites, for which estimates are required, must therefore be employed. This is an indirect approach to solving the problems, but must be invoked in cases where there are no local data on which to base flood estimates.

There are two principal methods of achieving these aims. One method involves a regional flood frequency analysis, the other requires the results from an analysis of rainfall and runoff behaviour to be translated from one catchment to another.

The regional flood frequency approach is based on the statistical similarity of observed flood behaviour in a region. It assumes that pattern of floods, both in terms of their frequency of occurrence, and their relative magnitude are similar, or homogeneous, throughout a region. In this sense, the study bears resemblance to the rainfall analysis described in Chapter 4. The study assumes that all catchment effects, such as slope, elevation, soil type, transmission losses, catchment area, etc., are included in the gauged flows.

Rainfall-runoff analysis starts by estimating a storm of certain rarity, areal extent, and duration, and then processes the rainfall generated by this storm to produce a runoff response from the catchment in question. In contrast to the flood frequency analysis, which only estimates peak flow, an estimate of the peak flow, flood hydrograph, and flood volume are generated. However, the procedure requires many subjective decisions in places where few data are available, such as Saudi Arabia. Further discussion comparing the two methods may be found in Appendix B.

Previous studies have used both of the methods of regional flood frequency and rainfall-runoff to derive estimates of floods at ungauged sites. However, in the light of the discussion in Appendix B, and as recommended by previous workers in this part of the

world (e.g. Walters, 1989; Wheater and Brown, 1989), we decided to focus our efforts on the flood frequency approach alone.

One of the fundamental drawbacks of the flood frequency method, as indicated above, is that it only provides information on peak flows: rainfall-runoff analysis produces a full hydrograph involving estimates of both the shape of the flood and its volume. If the rainfall-runoff approach is not to be used, then some alternative means of estimating the shape and volume of the flood must be found.

The approach we have adopted for the general flood frequency analysis includes the following steps:

collect and collate all available instantaneous peak flow data;define the standard index flood for each gauging station;define a homogeneous region which includes the Jeddah area;define a relationship for estimating the index flood at each gauging station from available

map data;standardise the flood series at each gauging station by the index flood derived from the

above relationship;establish the independence of floods at each station in the homogeneous zone;pool all the independent standardised flood events and treat as if from one record;define the flood frequency relationship from the pooled data and apply to the ungauged

catchments.

The method we adopted for establishing hydrograph shape and volume was based on the following steps:

estimate hydrograph shape following recommendations in published literature, based on catchment area;

calculate volumes under this hydrograph for each flood;use unpublished data on average, dimensionless hydrograph to estimate hydrograph shape;estimate flood volume from published relationship between flood peak flow and volume;scale dimensionless hydrograph to achieve estimated flood volume;compare resultant hydrographs from two methods and compare with IH (85) flood

estimates and volumes.

The unpublished data referred to above are relationships developed by the author for the Wadi Ghat, south of Jeddah, in the Yiba Basin. They are contained in an MSc thesis entitled ‘Arid zone rainfall-runoff modelling in south west Saudi Arabia: a case study’ (Brown, 1987). The published data referred to are contained in papers entitled ‘A method of estimating design flood hydrograph shape in an arid region’ (Walters, 1989) and ‘Limitations of design hydrographs in arid areas - an illustration from south west Saudi Arabia’ (Wheater and Brown, 1989).

5.4 Results

5.4.1 Peak Flows

A full description of the work undertaken may be found in Appendix B. The results contained in this section summarise the main points found in our analysis.

Data on instantaneous peak flows were available for the gauging stations indicated in Figure 5.1. Of these stations, those with data in which at least four years contained floods were collated. Full details of these stations and their flood statistics may be found in Appendix B.

The steps involved in the approach adopted for the study are outlined in Section 5.3, and are based on a regional flood frequency analysis.

The first task undertaken was that of processing the available peak flow data. This was not always so straight forward, in that, on many occasions, the recorded annual maximum instantaneous peak flows were less than the daily average noted for the day. In several instances these peak flows were not only less than the daily average flows for the day on which they occurred, but there were other days in the year which had higher daily flows. In these cases the years had to be dropped from the analysis as there was no way of reconciling the two sets of figures in the available time.

Having processed the data, we plotted all floods for each station, along with several fitted statistical distributions. The purpose of this was to define the most appropriate distribution that would consistently be capable of realistically describing the observed data. The distribution we finally selected was in keeping with that used by previous workers in this part of the world (e.g. IH, 1985; KAU, 1989). The parameters from the selected distribution for each station were then noted.

In order to define a homogeneous region we used both measures taken from the individually fitted distributions, as well as statistical measures estimated from the data. Previous workers have reported being able to identify homogeneous regions on the basis of these measures. Unfortunately, despite much searching, we could find no evidence distinguishing one part of the west coast of the kingdom from another.

We decided to use the flood with a five year return period of occurrence (Q5) as our standard. This measure is more meaningful than the mean annual flood, which is commonly used in other parts of the world, since many wadis do not flow at all in many years. IH (1985) used the same index for the same reason.

Since Q5 is the measure to be used in standardising the recorded floods at each station, it is also the measure which is required to be estimated at ungauged sites when the final results are applied. We therefore investigated its relationship to several different measures which we thought might be able to explain its variation from one catchment to another. The measures, or independent variables, included catchment area, length of the longest wadi course, average catchment elevation, average slope of the main wadi, latitude, and mean annual rainfall. In common with other workers we found catchment area to be the most important descriptor (e.g. IH, 1985; KAU, 1989). However, unlike other workers, we found elevation to be the only other descriptor that improved our estimates of Q5. Nevertheless, when we compared our estimated Q5 values with the original ones there

were large discrepancies.

We investigated these discrepancies by plotting the ratio of our estimated Q5 from the derived relationship with the original value. The results are shewn in Figure 5.2. There is a strong pattern revealed, indicating that Q5 is consistently being grossly overestimated by our relationship in the area around Jeddah, and being slightly underestimated for areas further south, where rainfall is higher. On this basis, we were able to define a homogeneous region around Jeddah. This region is indicated on Figure 5.2.

Having defined a homogeneous region, quite distinct from the area to the south, we redefined our relationship for estimating Q5 just using the stations in the region. The relationship developed includes only area, since, because it was based on just seven data points, it would be unwise to use more than one independent variable. The stations used to define this relationship are listed in Appendix B. The relationship is:

Q5 = 0.522 · Area 0.883

standard error = ×/÷ 2.03R2 = 0.78

where Q5 is the flood with a return period of five years (in m3s-1) and Area is the catchment area draining to the point of interest (km2).

Flood peaks for all stations in the homogeneous region were standardised using the value of Q5 derived from this equation. We checked for independence of events, working on the basis of independence of storm, rather than independence of weather system. The reason for adopting this meaning of independence was that, when considering the dates on which annual maximum floods occurred, we found that wadis separated by several hundred kilometres would have a peak flow on the same day, but gauged wadis in between would not.

Having standardised the floods from all gauging stations, and rejected the smaller of dependent annual maximum events, the data were pooled and plotted as a single series, in the same way as done for the regional rainfall analysis. A distribution was then fitted to the data. Figure 5.3 shews the results.

Since all the data have, by this stage, been standardised on Q5, the distribution represents the way in which factors of Q5 relate to the frequency of occurrence. For example, the 10 year flood is a factor of 1.76 times the Q5 flood. This ratio expressing the relationship of floods of different return periods to a standard flood is termed ‘the growth factor’, and the distribution in Figure 5.3 is called ‘regional growth curve’. The growth factors for several return periods of particular interest are provided in Table 5.1.

It should be noted that an adjustment of the growth curve was necessary to ensure that the five year return period flood had a growth factor of one. The nature of this adjustment is described in more detail in Appendix B.

These growth factors have been applied to all wadis in the study area in order to derive the required flood estimates for different return periods. The results are presented in Table 5.2. Also included at the bottom of the table are the flood estimates for the Wadi Rabigh

for comparison the Wadi Fatima. Rabigh is a large catchment in the Jeddah region, similar in scale to Wadi Fatima. The flood estimates shewn are derived from a distribution fitted to its own data. It should be noted that the station only has 15 years of data, so estimates of floods beyond a return period of 20 years are unlikely to be reliable.

5.4.2 Hydrograph shape and volume

Section 5.4.1 describes the results of the peak flow analysis. However, this tells us nothing about the shape and volume of the flood. The approach adopted for deriving the hydrograph shape and volume is outlined in Section 5.3. A more detailed description of the analysis is presented in Appendix B.

The design hydrograph procedure outlined by Walters (1989) derives from the Dames and Moore Representative Basins Study. It assumes that peak flow has already been estimated by some procedure, such flood frequency analysis, and bases the shape of the flood on relationships developed with catchment area. The relationships are described in Appendix B.

We applied this methodology to derive hydrographs for several of the catchments draining towards Jeddah, including Wadi Fatima. Although the method was not developed to estimate flood volume, an estimation of the hydrograph shape implicitly estimates the flood volume as well. We therefore derived the flood volumes under the hydrographs for these catchments.

We also used a relationship developed for the Wadi Ghat, in the Yiba Basin, south of Jeddah, which was based on an analysis of 11 floods. The relationship has been suggested as offering a way forward for estimating flood volumes (Wheater and Brown, 1989). The background to the relationship is described in a little more detail in Appendix B, but the equation developed is:

Vol = 7.07 Qpk

standard error = +/- 176MlR2 = 0.84

where Vol is the estimated volume of the flood (Megalitres, Ml), and Qpk is the peak flow of the flood in question (m3s-1).

This relationship relies upon the shape of the hydrographs from which it was derived. These hydrographs have characteristics which are noted as common to all arid zone hydrographs, such as steep rise times, sharp peaks, and fast recessions down to a diminishing baseflow. Not surprisingly, therefore, the bulk of the volume of the flood is associated with the peak flow, which is exactly what this relationship demonstrates. There is, therefore, no reason why it should not be applicable to any similar shaped hydrograph from an arid zone. Unfortunately we did not have any data to verify the relationship in the Jeddah region, but given the commonality of hydrograph shapes experienced from arid countries all over the world, we felt it was justifiable to use the relationship in this study.

We compared the results of the volumes predicted from this relationship with those derived from the Dames and Moore hydrographs described above. Compared to the flood peak-volume relationship, the Dames and Moore hydrograph implied volume was underestimating for small catchments, and overestimating for large catchments. When the hydrographs were examined visually, is was clear that non-standard shapes were being predicted. This was a consequence of the methods reliance for hydrograph shape solely on catchment area. The catchments for which unusual shapes were being predicted had catchment areas outside the range of those used to develop the methodology. This included both the small wadis (less than 100km2) as well as Wadi Fatima.

We therefore needed a more robust method of defining hydrograph shape that could also provide volumes consistent with those derived from the above relationship.

In the study of Wadi Ghat runoff (Brown, 1987) an average unit hydrograph response was developed. Whilst it was noted that it did not replicate flood events particularly well, this was for reasons of scale and timing, rather than shape. In other words, if the timing of a flood after rainfall is unimportant, and the scale of the flood peak is already known, then this unit hydrograph has the potential to provide a robust shape, scaled to both flood peak, and runoff volume. We therefore standardised this unit hydrograph on peak flow and time to peak to produce a dimensionless hydrograph, whose ordinates are presented in Table 5.3.

Time (% of time to peak) Flow (% of peak flow)

0 0

10 20

20 60

40 75

100 100

140 90

170 75

230 60

300 45

400 30

600 15

750 10

900 7

1100 5

1500 2

2000 0

Table 5.3 Dimensionless hydrograph ordinates

The time-to-peak estimate needed to redimensionalise the hydrograph may be calculated iteratively to derive the required volume. When using the above peak flow-volume relationship, this is achieved with a time-to-peak of 33 minutes. Although, at first site, it would seem unreasonable to expect floods of all sizes and from all catchments to have the same time-to-peak, Walters (1989) noted that there was no relationship with catchment size (which, for his data, varied from 170km2 to 2672km2) or any other variable that was considered. Time-to-peak was found to vary from 12 to 84 minutes, with a mean of 32 minutes. He recommends either selecting an arbitrary time-to-peak estimate, or using a mean value of 30 minutes for design purposes. The agreement between his mean design estimates for time-to-peak, and that required in order to match the flood volumes of our dimensionless hydrograph and peak-volume relationship, is, therefore, excellent.

We therefore based the design hydrographs and runoff volumes, for all wadis in the study area with catchments greater than 10km2, on the peak-volume relationship and dimensionless hydrograph described above. As an illustration of the results, the design flood hydrographs for the Mushwab Dam catchment are shewn in Figure 5.4

5.5 Discussion

5.5.1 Peak flows

There are two key components to the regional flood frequency analysis. These are the estimation of the index flood, Q5 in this case, and the shape of the derived growth curve.

The equation derived for estimating Q5 by IH (1985) was based on data from 17 catchments along the west coast of the kingdom. Their equation was:

Q5 = 2.818 · Area 0.72

for which they quote a correlation coefficient of 0.92. No standard error is mentioned. Our comparable equation is:

Q5 = 0.522 · Area 0.883

It is plain to see that, particularly for smaller catchments, the estimate of Q5 from the IH equation will be substantially larger than that estimated from our equation. For example, the Mushwab Dam catchment has an area of 37.9km2. The estimates for Q5 using the two

equations are 38.6m3s-1 and 12.9m3s-1, respectively. This factor of three in the difference between the two estimates requires an explanation.

The reason for the difference between the two equations lies in the data on which they were based. The IH relationship was a very strong one (R2 = 0.92) and based on 17 stations. However, the majority of these were located in the south west of the kingdom, with only two stations from the Jeddah region.

Our regional relationship for estimating Q5 was much weaker than IH’s, primarily because we included the six available gauging stations in the Jeddah region. This prompted us to look into regional variations in the estimation of Q5, with the discovery that Q5 in the Jeddah region was consistently being grossly overestimated by the regional equation. We therefore redefined our relationship for Q5, with this in mind, specifically for the Jeddah region using only those gauging stations in the Jeddah region indicated in the analysis of residuals (see Figure 5.2).

It is therefore not surprising that the regional equation developed by IH produces much higher estimates of Q5 in the Jeddah region than our local equation.

The regional flood frequency growth curves can be described in terms of their growth factors at different return periods. Table 5.4 presents a selection of growth factors for both the IH (1985) curve and our curve.

Return Period (yrs) 5 10 20 50 100IH (1985) 1.00 1.64 2.36 3.56 4.52WSA (1994) 1.00 1.76 2.87 5.20 7.97

Table 5.4 Comparison of growth factors from IH (1985) with those derived in this study for the Jeddah region.

From the table it can be seen that the growth curve derived in our study is substantially steeper than the IH curve. The effect of this is not so noticeable for small catchments. For example, continuing to use the Mushwab Dam catchment, the one hundred year return period flood peak (Q100) would be estimated using the IH method as 174m3s-1. Our estimate of the same flood peak is 103m3s-1. The difference between the Q5 estimations is still the dominating factor.

However, for larger catchments, the difference does make itself felt. For example, the Wadi Fatima has a catchment of 4597km2. The IH estimates of the Q5 and Q100 peak flows are 1222m3s-1 and 5522m3s-1, respectively. Our estimates for the same peak flows are 895m3s-1 and 7134m3s-1, respectively.

Once again the reasons for the differences lie in the data sets used to derive the curves. Our data set is based solely on those stations that lie within the region defined in Figure 5.2. The IH analysis was based on all stations along the coastal strip. In our search for a homogeneous zone, nothing was indicated by the data that the distribution of floods in any one part of the south west was significantly different from that displayed in other part of the region. Whilst it would appear, therefore, that none of the stations in the Jeddah region, alone, had sufficient data to display regional idiosyncrasies, the station year

analysis of flows suggests that, like the rainfall, flows in the region around Jeddah are generally lower, but more variable, than those elsewhere along the coast.

5.5.2 Hydrograph shape and volume

Previous workers have used many different types of hydrograph shape to define runoff response. In the arid zone this has included the use of synthetic hydrographs, triangular hydrographs and those determined from catchment parameters, among others. They have mainly been shewn to be unreliable, and unrepresentative of the typical response observed in arid regions (e.g. Brown, 1987; KAU, 1989).

In this study we have looked at two alternative methods for design hydrograph estimation: the Dames and Moore method (Walters, 1989), and our own method developed from work done by Brown (1987).

For catchments whose areas fall within the range used to develop the method, the Dames and Moore technique produces believable results. Figure 5.5 shews the design family of hydrographs for the Bani Malik catchment, which drains an area of 278km2, using a time-to-peak of 30 minutes. The hydrographs have a reasonable shape, with fairly steep rising limbs, a sharp peak, and a relatively rapid fall down to medium flows. The recession has been adjusted from the recommended design, which suggests an exponential decay. The hydrograph is somewhat angular, but for design purposes is entirely adequate.

However, for smaller catchments, the parameters used to define the shape of the hydrograph become unrealistic. Figure 5.6 shews the most realistic Dames and Moore response (using a 5 minute rise time) compared with the WSA response derived using the dimensionless hydrograph in Table 5.3. It can be see that the falling limb of the hydrograph is exceptionally steep, with a sudden break in slope when flow reaches a quarter of the peak flow. Using the original exponential decay function would make this break even more pronounced as its gradient is much shallower. Using a longer time-to-peak causes the falling limb to be vertical, or even sub-vertical. It can also be seen that the volumes under the two hydrographs are different by a factor of around two. The volume under the WSA hydrograph has been calibrated by adjusting the time-to-peak so that it coincides with the volume predicted by the peak-volume relationship in Section 5.4. In this case, the WSA hydrograph produces the more realistic results.

Moving on to larger catchments, Figure 5.7 compares the WSA design hydrograph with that from Dames and Moore. Here it can be seen that the large catchment area has resulted predicting an excessive width for the hydrograph. This kind of hydrograph shape is occasionally seen in arid countries, but is usually associated with relatively small flows, where the residual baseflow following the flood event is significant. This shape would not be expected for a 7000m3s-1 flood. The volumes under the two hydrographs are different by a factor of around two, as in the previous example, but with the Dames and Moore volume being larger in this case. Once again, the WSA design hydrograph provides more realistic results.

Directly comparing these hydrographs and volumes with those derived by IH (1985) is not possible for the catchments draining in to Jeddah. This is because the catchments for

which they derived runoff hydrographs have different areas than those in our study. Our flood estimates are required for sites further up the wadis. However, the results for the Wadi Fatima are comparable, because that catchment area has not changed.

The volume of the 100 year flood estimated by IH for the natural Wadi Fatima catchment to its end is 328,300Ml. This compares to the volume under the Dames and Moore hydrograph of 106,000Ml, and our estimate of 50,000Ml. Thus, there is a difference of 6.5 between our estimate of the volume of the Q100 flood volume and that of IH.

The reason for this difference can be seen in Figure 5.8, which compares the calculated runoff hydrograph for Wadi Fatima from the IH rainfall-runoff analysis with the design hydrographs of Dames and Moore and WSA. The very large discrepancy between the two sets of hydrographs results from the estimation of design storm duration in the IH method. The time-to-peak of the unit hydrograph was multiplied by a factor of 12 to derive storm duration. However, the basis for using this factor is not explained: the report merely states that the relationship “gave reasonable durations for the range of catchments studied.”

Comparing these hydrographs with other published hydrographs, and personal experience in the field of arid zone hydrograph analysis, suggests that the WSA design hydrograph for Wadi Fatima, in Figure 5.7, is the most reasonable, and that of IH in Figure 5.8 is the least reasonable. For example, in Walters (1989), the hydrograph of a gauged 3,200m3s-1 flood is presented. The flood starts, rises steeply, peaks sharply, recedes rapidly, and finishes in 8 hours. This is a very similar shape and duration suggested by the WSA design hydrograph of 10 hours, and supports the design which we recommend for adoption.

6 Summary and Conclusions

6.1 Rainfall

We have updated previous rainfall depth-duration-frequency curves derived for design purposes in Jeddah using the station year method. Our results fall midway between those of the previous Gibb studies (1982, and 1985). The reasons for differences can be summarised:

we have eight raingauges with similar rainfall characteristics to Jeddah than the three that were used in the Gibb 1982 study;

we have focused on analysing raingauges shewn to be homogeneous with rainfall in Jeddah, rather than using all raingauges within a radius of the city (the Gibb 1985 approach);

rainfall along the coastal plain, while being less, on average, than that in the jebel, is more variable than found in the jebel;

the standard rainfall for Jeddah, from which all other rainfall depths are calculated, is around 10% higher in our study than in the Gibb (1985) study.

The results of our analysis for design use are presented in Table 4.2 and Figure 4.4. Rainfall depths are given for Jeddah and have been derived by multiplying the appropriate factors by the index rainfall for Jeddah (41.6mm, the depth of rainfall exceeded in one hour once every five years, 1H-M5). The factors presented in the table may used anywhere along the west coast of Saudi Arabia between Jeddah and Jizan, for sites below 100m above sea level, by multiplying by a local estimate of the 1H-M5 rainfall derived from published data, or from a local raingauge.

We have analysed the available information on storm profiles. This included data from the MEPA raingauge in Jeddah and five other raingauges located in the Wadi Ghat catchment in the jebel to the south of Jeddah. We are unaware of any other previous analysis on the shapes of storms along the west coast of the kingdom. The principal findings may be summarised:

two thirds of storms less than three hours long appear to have intense rainfall at the beginning of the event, with 70% of the total storm depth falling in the first 40% of the storm's duration;

one third of storms less than three hours long commence with an extended period of less intense rainfall, 60% of the total storm depth falling in the last 40% of the storm duration;

most storms lasting longer than three hours have heavier rainfall in the first half of the storm, although they may occasionally take the reverse of this profile;

Our storm profiles recommended for design use are presented in Table 4.3, and Figures 4.6, 4.7, and 4.8. They may be used by multiplying the time percentages in the appropriate section of the table by the desired storm duration (in minutes). The depths at each

corresponding time interval may be derived by multiplying the appropriate depth percentages by the total storm depth derived from Table 4.2 for the desired duration.

6.2 Wadi Flows

We have analysed the characteristics of wadi flood flows around using a regional station year approach. Our results indicate smaller, but more variable, peak flows are experienced by most wadis than predicted in the Gibb (1985) study. While we think peak flows on the Wadi Fatima are smaller for more common events, we estimate floods of more rarity will have bigger peak flows. We also suggest flood volumes are substantially less than proposed in the Gibb study. The main reasons for the differences are:

more wadi flow gauging stations in the Jeddah region were available to us than to the Institute of Hydrology (IH) in the 1985 Gibb report;

with the additional data, we found that wadis around Jeddah display a different relationship with catchment area than those further south, used by IH;

we also found that, when combined to represent a longer time period, floods around Jeddah, like the rainfall, are more variable than those in higher rainfall areas further south along the coast;

we have based our estimates of flood volumes on a previously established relationship with peak flows developed for a wadi in the region, rather than the tentative rainfall-runoff procedure used by IH;

our runoff hydrograph is an average shape response developed on a catchment to the south of Jeddah, rather than standard a triangular hydrograph.

Our recommended relationship for estimating the five year return period index flood (in m3s-1) is:

Q5 = 0.522 · Area 0.883

standard error = / 2.03 R2 = 0.78

The relatively large standard error should be noted. This indicates the degree of confidence that can be placed in the accuracy of the relationship. Given the variability in wadi floods, the difficulty of gauging wadi flows, and the relatively short lengths of gauging records, this figure is not unacceptably high. The relationship may be used for any catchment on the west side of the escarpment between latitudes *** and ***.

Peak flows (in m3s-1) for selected return periods may be estimated by multiplying Q5 derived from the above relationship by the appropriate factor in Table 5.1. Alternatively, the appropriate factor may be read off from the curve in Figure 5.3.

The flood volume (in Megalitres, Ml) may be estimated from the following relationship with peak flow:

Vol = 7.07 Qpk

standard error = +/- 176MlR2 = 0.84

The design hydrograph may be derived from Table 5.3 by multiplying the flow percentages by the estimated peak flow. The time axis may be dimensionalised by multiplying the time percentages by 33 minutes, and checking that the volume under the derived hydrograph corresponds to that estimated from the above equation.

7 Recommendations

7.1 Recommendations for Design

Our recommendations for adoption in the hydrological design of the storm drainage system for Jeddah are summarised in Chapter 6. For rainfall, these include:

updated storm depth-duration-frequency curves;an updated value for the standard index rainfall, the one hour five year return period (1H-

M5) rainfall depth;new storm profiles based on observed rainfall intensities.

For wadi flows, our recommendations include:

an updated relationship for deriving the standard index flood, the five year return period peak flow, Q5;

an updated regional flood growth curve;a new relationship for estimating flood volume;a new dimensionless hydrograph for defining flood shape.

We believe these relationships make the best use of the available data, and until such time as more data are available, supersede all previous work and provide the most reliable estimates of hydrological variables, for design purposes, in and around Jeddah.

7.2 Recommendations for Further Work

With the presently available data, we do not believe that much more information could be extracted. There are certain parts of the analysis where more work might yield more confidence in design criteria, such as the analysis of wadi flow hydrographs, and flood volume-peak flow relationships. However, these are research exercises, and, with the difficulties of obtaining the appropriate data, could not be undertaken in the timescale of this project.

However, there are certain gaps in the original data which impact directly on the degree of confidence that can be placed in the recommended design criteria. These relate both the wadi flows and to rainfall.

As previously indicated in Section 4.4.1, the variability of rainfall over the city of Jeddah, or, for that matter, over any part of Saudi Arabia, is largely unmeasured, and therefore unquantified, and unknown. Dense networks of raingauges have been installed in research catchments but failed to define areal storm profiles. Denser networks are uneconomic to install and maintain, and have limited areal application. The frequency of storms over given areas is therefore largely determined by the chance of a storm being recorded by any given raingauge in the existing network. These unknowns constitute a major gap in the

present knowledge of rainfall in Saudi Arabia in general, and in Jeddah in particular. Until this gap is filled there is little likelihood of developing a standard approach to rainfall-runoff relationships in the kingdom for design or water resources purposes. In order to address this major gap in the knowledge of rainfall variability and distribution, not only for Jeddah, but also for wider understanding and application, we therefore recommend that

the installation of a weather radar post overlooking Jeddah should be seriously considered.

Turning to wadi flows, several elements combine to compound the uncertainty in flood estimation in the wadis draining towards Jeddah. These include:

none of the wadis flowing towards Jeddah are gauged, including the Wadi Fatima;the lack of adequate rainfall description means that rainfall-runoff relationships are too

indeterminate to provide any reliable estimates of flood flows;flow calibration at all gauging stations in the kingdom appears to have been cessated since

1987.

This last point is particularly serious since it is a consequence of the withdrawal of financial support from field hydrology activities. It indicates that there is a lack of commitment to any long term outlook concerning hydrometric data collection. Given the variability, both in time and space, of flood events, and of their sometimes catastrophic consequences, this outlook appears to be singularly short-sighted. We understand that this does not fall under the remit of MOMRA, but the consequences of the dereliction of the wadi flow gauging network in the Kingdom include an unhappy prognosis for those that choose to dwell, in particular, in major wadis, such as the southern end of Wadi Fatima. With this in mind, and in conjunction with the above recommendation for the installation of a weather radar, we make the following recommendations:

a strategic assessment is undertaken to establish a long term plan for the future of the hydrometric network in the kingdom;

a flow gauging station on the Wadi Fatima is installed, for example, at the Jeddah to Makkah road bridge;

at least two other gauging stations are installed on the wadis flowing towards Jeddah: one of these should be on the Wadi Ghalil at the culvert under the Makkah road; the other should be on another well defined wadi to the north, such as Wadi Dighbij;

a modelling study is undertaken on the Wadi Fatima to assess the depth, direction, and volumes of the one hundred year flow across the coastal plain.

This latter recommendation arises as a consequence of the very large uncertainty surrounding the course of the Wadi Fatima as it exits onto the coastal plain. Much of the coastal plain at the exit of the wadi is indicated as potential development area in the town plans for Jeddah. Given the estimates of peak flow and volumes of water involved, and the extent of planned development, the requirement to undertake some investigation into the

likely behaviour of a major flood from this wadi is compelling. We therefore recommend that the study should include the following components:

detailed topographic survey of the mouth of Wadi Fatima and the area of the coastal plain around its exit, to the coast;

detailed cross-sectional information on the wadi course from a fixed upstream point, such as the Makkah road bridge, to its mouth;

sand/soil profiles at strategic cross-sections down the wadi from the Makkah road bridge to the coast for the purposes of modelling depth of scour;

building a one-dimensional hydraulic model of the wadi from the Makkah road bridge to the mouth of the wadi;

building a two-dimensional hydraulic model of the coastal plain, linked to the 1-d model;

both models to incorporate scouring processes relevant to wadi flows, and sediment transport processes;

a phased approach to model building and data collection, using initial, coarse models to indicate where detailed data collection is required.

We would be pleases to draw up the terms of reference for such a study, should this be required.

ReferencesWan