Suitability Of Commonly Used Design Storm Profiles For Oman

Models relying on rainfall runoff transformation require, as input, design storm hyetographs that are temporal distributions of the design rainfall. Veneziano and Villani (1999) grouped commonly used design storm development methods into four major categories: (1) geometrical forms anchored to the IDF curve at a single point; (2) the Alternating Block Method (ABM) (Chow et al., 1988) based on the full IDF information; (3) developing standardized profiles directly from raw rainfall records; and (4) stochastic-model simulations (Cunderlik and Simonovic, 2004).

Recognized methods of producing standardized profiles from raw rainfall records (category 3) include those of Huff (1967), the four classes of storm profiles presented by the SCS (currently National Resources Conservation Service) (Kent, 1973), and the Flood Studies Report (NERC 1975) method used to derive the UK storm profiles. Huff (1967) collected information on 261 storms in the USA, classified the storms into four quartiles – according to when each storm’s rainfall intensity peaked – and derived dimensionless hyetograph curves corresponding to the 25^th, 50^th, 75^th, and 90^th percentiles.

Four types of dimensionless hyetograph curves (types I, II, III, and IA) were developed by the Soil Conservation Service (SCS) corresponding to storms from across the United States (Kent, 1973). The cumulative curve – which is the basis for type II distribution for example – was “established by (1) plotting a ratio of rainfall amount for any duration to the 24-hour amount against duration for a number of locations and (2) selecting a curve of best fit.” These best fit cumulative curves were derived from average 30-minute intensity-duration values. The greatest 30-minute depth was placed in the middle of the 24-hour period and would be bordered on either side by the second and third largest depths (30 minutes after and 30 minutes before). The method is essentially the Chow et al. (1988) Alternating Block Method (ABM) procedure. All durations are nested within one single design storm profile. Hence, multiple runs are not needed to determine which storm duration is the most critical one for a specific stormwater network design/simulation.

Recently, Chin and Ross (2018) and Chin (2020) restudied the SCS storm profiles and advised that the SCS type II storm profile region of application in the USA should be further split into three regions with their corresponding storm profiles categories. The new storm profiles are having more conservative sub-daily ratios than those of the SCS type II. It is worth mentioning that the nested storm profiles (via ABM) could tend to overestimate the intensities of storm events, particularly for frequent events (Frederick et al., 1977). Nevertheless, ABM methods are common in codes of practice. They feature, for example, in most drainage manuals used across the USA, in the HEC-HMS package Technical Reference Manual (Feldman, 2000) and relatively new handbooks and textbooks such as Chin et al. (2013).

Fewer studies have focused on the Arab Gulf regions. Awadallah and Younan (2012), Elfeki et al. (2014), Ewea et al. (2017) and Awadallah et al. (2017) advised against adopting the SCS type II and/or the UK50% Summer storm profiles (which are the most conservative storm profiles among SCS and UK, respectively). For Oman, Al-Rawas and Valeo (2009) derived storm profiles based on hyetograph recordings. However, they confused the processed form of data with that of the original hyetographs and thus obtained front-loaded design storm profiles, with peaks in the very beginning of the storm events. Awadallah et al. (2107) derived four storm profiles for Oman; however, the data were limited to 236 storms.

A scan of codes of practices used in Arab countries showed that Egyptian (MWRI, 2011) and Saudi Arabian (Riyadh Municipality, 2012; MOC, 1989) codes of practice stipulate the use of the SCS type II profile; while codes of practice in Qatar (PWA, 2005), Kuwait (Hyder Consulting, 2002) and Dubai (MWH, 2014) recommend the use of the UK50 summer storm profile. All of these codes used the SCS type II and UK50 with their corresponding American- or British-calibrated values without considering recalibration using actual rainfall records collected locally. The Omani Highway Design Manual issued by Oman’s Ministry of Transport and Communications (2010) is the only code of practice that proposes a less commonly known profile (that of Wheater and Bell, 1983), which is similar but more conservative than the original SCS type II profile.

This paper aims to investigate the safety and appropriateness of using dimensionless design storm hyetographs in hyper arid and arid regions, using recorded hyetographs for storms that took place in Oman. The data were obtained from 48 rainfall stations distributed across Oman. A design storm profile is derived from the recorded hyetographs. The obtained results were compared with the SCS type II, the newly developed Chin (2020) SCS-type curves, and the Wheater and Bell (1983) profiles.

For this research, data for 5895 storms were obtained from the Ministry of Regional Municipalities and Water Resources (MRMWR), Oman, in various previous projects of Dar Al-Handasah. Data were collected from 48 rainfall gauges located across North-Eastern Oman. The rainfall records used in this study date back to various years, ranging from 1974 to 2010. However, data with available hyetographs, with data at a time step of 5[thinsp]min or more, extends back to 1993 only. The map in Figure 1 illustrates the locations of the rainfall gauges, while Table 1 presents metadata characteristics for each station. Storms with durations less than 10 minutes or with total depths less than 10[thinsp]mm were excluded from further analysis, and the total number of storm events left is 1354 storms.

As described in the literature review, design storm profiles can be derived through several possible methods. However, all methods start by identifying prevailing storm patterns. Storms were thus classified based on their durations and storm profiles were investigated to check if they have random patterns or distinct characteristics, such as peak position with respect to the total storm or ratios of peak intensities to average intensities. If the analysis does not identify any such patterns, design considerations will override meteorological considerations. In other words, it would be wiser to derive storm profiles that produce the highest possible peak discharges, bearing in mind the common rainfall-runoff models.

The 1354 storms (with durations exceeding 10[thinsp]min and rainfall depths exceeding 10[thinsp]mm) were classified into six categories according to their durations: Category 1 (10 minutes – 1 hour – 264 storms); Category 2 (1 hour – 3 hours – 414 storms); Category 3 (3 – 6 hours – 237 storms); Category 4 (6 – 12 hours – 241 storms); Category 5 (12 – 24 hours – 154 storms) and finally category 6 (storms exceeding 24 hours - 44 storms). For each category, dimensionless cumulative hyetographs were drawn between the dimensionless cumulative duration (i.e., percentage of the cumulative duration to the total storm duration), on the x-axis, and the dimensionless cumulative rainfall depth (i.e., percentage of the cumulative rainfall depth to the total rainfall depth), on the y-axis. Figure 2 shows the cumulative dimensionless hyetographs for the six categories. The results demonstrate that peak intensities can occur indiscriminately throughout the duration of the storm. This variation makes it even more important to search for a design (critical) storm that could produce higher runoff discharges, with symmetrical, nested, peak centred profiles, like those derived by the SCS method, using ABM.

To investigate the safety of using the SCS derived profiles, the same procedure used in Kent (1973) is applied on the 1354 available storms. This method comprises two steps. First, rainfall depths are calculated for different storm durations, using a user-defined analysis time step generally ranging from five to ten minutes. Next, the incremental rainfall depth blocks are alternately arranged on either side of the centre of the storm hyetograph (Figure 3) and the individual hyetographs are averaged to obtain a SCS type profile, however based on the Omani data and not the USA data. We term this derived profile “Calibrated SCS curve for Oman”.

From Figure 3, most of the data cloud (shown in grey) is much higher (more conservative) that the SCS type II standard curve (shown in green). This data representation allows to plot the individual hyetographs on the same plot. The calibrated SCS curve for Oman (shown in red) is more centred within the Omani data cloud of individual hyetographs and thus is more representative of the Omani storms. It has a sharper peak, which reflects the nature of Omani storms with more than 65% of the storms (with durations exceeding 10[thinsp]min and rainfall depths exceeding 10[thinsp]mm) of less than 6 hours.

The 24-hr dimensionless rainfall profile developed using the nested ABM approach of the SCS were compared to the standard SCS type II profile (the most conservative among standard SCS types), the one developed by Wheater and Bell (1983) and the newly developed profiles by Chin (2020). This is another way to represent the nested design profiles, more common than the representation of Figure 3.

All the above mentioned nested, peak-centred, storm profiles are plotted once for the entire 24-hr storm duration and with a zoom in the middle 2 hours for better comparison (Figure 4). The developed profiles showed steeper and more critical rises than the SCS type II, which indicates that the latter profile (and hence all the other SCS profiles which are less conservative) is inadequate for use in Oman and probably in other arid regions. On the other hand, the developed storm profiles showed less critical rises, compared to the Wheater and Bell (1983) storm profile, which is the one stipulated for use in the Omani Highway Design Standards (2010). The closest previously published storm profile to the one developed is the Chin (2020) Cluster 1 storm profile, representative of the relatively arid and semi-arid Central Plains and Southwest, USA, which is the most conservative storm profile in the USA, according to Chin (2020) study.

Using raw rainfall hyetograph data from the Ministry of Regional Municipalities and Water Resources of Sultanate of Oman, coming from 48 gauging stations recording over the period 1993-2010, 1354 storm events (with durations and depths exceeding 10[thinsp]min and 10[thinsp]mm, respectively) were analysed to investigate if special patterns are present in the raw storm hyetographs and then are used to develop representative and data-based design storm profiles for the region. The Alternating Block Method, like the SCS approach, was used to develop a nested 24-hr storm profile. From the outcome of this research, a few recommendations could be drawn:

• The studied Omani storm hyetographs vary in duration, shape, peak-timing, and occurrence, with no detectable patterns. Consequently, design considerations rather than meteorological ones should guide the development of design storm profiles, to generate safe peak discharges when used in rainfall-runoff models.

• A 24-hr nested storm profile was developed in the framework of the current study, based on the SCS approach. The developed profile is more conservative (result in higher peak discharges) than the SCS type II, even though the latter profile is widely used in many codes of practice in arid and hyper arid regions.

• The nearest profile from published literature to the developed one is the SCS-type Cluster 1 storm profile, developed by Chin (2020). The Wheater and Bell (1983) storm profile, stipulated in Omani Highway Design Standards (2010), is more conservative than the developed storm profile; and hence is expected to produce a more conservative (and potentially more costly) design.

• Finally, the study should be extended using data from other arid and hyper arid countries to confirm and generalize the results obtained for Oman in this study.

The authors acknowledge the tremendous support from the MRMWR, Oman, in providing the raw rainfall data.