a ‘THE DISTRIBUTION OF MONTHLY RAINFALL INTENSITY AT SOME SITES IN SAUDI ARABIA A.M. ABOUAMMOH® Diseurion of Rana Sal Arabir (Received March 1930) Abdeact. The anal of rainfall intensity is sein various fell oF fees agrculturl planning ydrology ‘and transmision of microwaves and high-voltage electricity. The monthly precipitation total fora 2 year [eriodare used ta compare the runfl regimes at seven ites im Saudi Arabi. These sitesi inher aide, Tongitude andclevation above sea level. Some asic monthly statistics of data fom hese sesare presented to idem the nature ofthe rainfall a each ste The read of ce number of dy months per yea iv also used ‘compare between est, The probability of dry months for each month of the 2 yrs wed for sompurson between the rainfall regims, Plas fF mean and maimmum fainfal a hese Sites are presented. The Fisher ‘Cornish proposed model for xinfllinari regions i considered aaa simple empirical method for estimating its parametes is used for the twelve month period rainfall dat rom the seven sits 1. Introduction “The Kingdom of Saudi Arabia is mainly an arid country with an estimated population of over |2 milion in an area of about 2156000 km", The area iactudes about three fourths of the Arabian peninsula, (Hamza, 1968: Tayeb, 1978). In the past fifteen years the country has seen the establishment of large agricultural projects. Current agricultural production hhas reached 4,6 million tons of which over 2.5 million tons was wheat in 1986 compared to about 3900 tons in 1972. Ground water is the main water source for irrigating the majority of these agricultural projects as well as forall water authorities in various cities, cowns anid villages. The ground water level above the water table atany site depends upon the amount of precipitation through the surface runoff intensity and the infiltration capacity (the passage of water through the sol), (Linsay et af, 1982), To fact, detaited analysis of rainfall intensities is useful in diferent aspects of earth sciences. Rainfall intensity determines the contribution of rainfall to the water budget, the sol profile, the «ecological niche and the watershed. Revently, in 1986, the Saud National Commission for Wildlife Protection and Development was established. Regions of Saudi Arabia which are put under the management of the Commission for wildlife devefopment are expected to have relatively higher monthiy rainfall ‘Therefore, itis of great importance to researchers in agricultural planning and forest ‘management to study the longterms runoff of rainfall and its effect on water levels In fact rainfall which is used isthe only input that varies, [rom Year to year, toestimate the Water balance or the crop yields (Stera anc Coe, 1982), even in countries of arid nature Data analysis of rainfall intensity is also Useful to researchers in electrical engineering + Postal adres: Department of Stats, College of Se, King Sat Univesity, PO, Hos 2888, Risad 11451 Sd Ara Freirnmental Mororng amd Assessment F-¥9-10, 1991 {©1991 Kluwer Academic Pubes. Pra nthe Neverland,sory ALM. AROUAMMON Its attenuating effect on microwave transmission in telecommunications has been considered by Bodimann and Ruthroft (1976) and Drufuca (1977). Its connection to power oss along high-voltage transmission lines, generating audible and tadio noise. has ‘been studied by Krikham (1980) Rainfall intensity (mm hr, month or year) and duration or frequency of rain (in hours ‘or days y") are the major Features used to characterize rainfall. In This study. we are concerned with the accumulated total rainfaiiin mm month-* rather than the time scale in ‘mm yr. This emphasis is based on the available data and the arid nature ofthe considered regions which have relatively small monthly accumtlated rainfall totals, “The main theme of this paper is to present an overall estimate and comparison of the rainfall regimes at seven different sites in Saudi Arabia. These sites, located in different provinces in Saudi Arabia are Dhahran, Jeddah, Khamis Mushait, Madina, Riyadh, Tabouk and Taif. Their latitude, longitude and elevation range from 818” to 28°22", 36° 35” to 50 10" and 17 to 2087 m, respectively, (Table 1). ‘The paper considers monthly rather than the daly rainfall since most daily rainfall records are zeros. Thus daily do not give reasonable background information for most inferential procedures ‘The sequel of the paper's the following, Some basic statistics of the monthly rainfal\for the seven sites are discussed in Section 2. In Section 3, the probability of no rain in each month and the probability of dry months in the year arc plotted in two figures for each of the seven meteorological sites. In addition figures for the number of dry months in the year, the maximum amount of rainfall and the mean amount of rainfall are used for possible comparison between the rainfall regimes of these sites. In Section 4, the Fisher- Cornish proposed model, (Fisher and Cornish, 1960), usedasa probability distribution to rainfall data with zero observations, is considered. Ozturk’s method (1984) is used 10 estimate the model parameter and finally in Section 5 some concluding remarks and ‘comments presented. 2. Data and Basic Statistics Daily rainfall data are provided for the seven meteorological stations Dhahran, Jeddah, Khamis Mushait, Madina, Riyadh, Tabouk and Taif by the Meteorology and Environ- mental Protection Administration ofthe Ministry of Defence and Aviation, Saudi Arabia. {A 2L-yr record from 1966 10 1986 is used for this study. Data from other stationsare not available at present for a sufficient number of years. Table I presents the number of rainy months at the seven location. Table III includes some basic statistics of the monthly precipitation rainfall, the mean ¥ and the standard deviation s° are presented. The ‘minimum monthly rainfall are zero except for May in Khamis Mushait (2.9 mm). ‘The data indicates differences between theseven sites with respect to the annual amount of precipitation during the 2-yr period (1966-1986). Khamis Mushait has the maximum amount of annual mean rainfall (38.23 mm in May). Taif has the next largest amount of ‘annual rainfallin April(37.76mm). The maximum totalamount of rainfall during the 21 year (1966 to 1986) is in Khamis Mushait (2494.68 mm) whereas the total amount ofDISTRIBUTION OF RARNEALI IN SAUDE ARABIA [sis TABLE Latitude, longitude and eeatian af even meteorological tes Latitude Longitude site sentiwte lesaton = Deg Min Deg Min oo Dhahran 2% 16 % 0 v Jeddah a wo » 2 ” Khamis Mushait 18 2 “6 ast Madina u a » a 6 Riyadh u 2 ea “ on abou as a % 3 76 Tif 2 » o 2 vasa TABLE Number of rainy months fr each month at the seven actions Locations Ce ee ee Dhahran Jeddah Khamis Madina “Riyadh Tabouk Tal, Musa fan ” a 1B 0 16 2 2 Feo. ts 3 0 6 te w 0 Mar » 1 1% 2 2» 6 ” May 5 5 a " 16 3 0 Jun 0 ° 16 3 : ° 2 a 0 t 2 t 2 ° 4 ‘Aue » o Is 4 2 t u Sep 0 t 0 3 1 o 6 oe, 3 9 5 i s Bb Now. 0 1 2 2 5 B 16 6 0 0 9 2 Dee 2 I ‘ainfalls in Taif and in Riyadh ate 2016.48 and 1274.04 mm, respectively. ‘The arid nature is apparent even in Khamis Mushait where the minimum monthly rainfall was zer0 forall months in the 2l-yr period except for May which hada minimum rainfall of 29 mm in 1972. Dry months with no rainfall from 1966 to 1986 are from June to September in Dhahran, June and August in Jeddah and June, July and September in Tabouk. Months with rainfall less than 5 mm are: May in Dhahtan, March, July, September and October in Jeddah, August in Madina, June and September in Riyadh and August in Tabouk ‘The probabitty of rainfalls greater than 0.85, (based on 1966-1986 data) in the period January to Mare for Dhahran, March to Mayand August for Khamis Mushait, Aprilfor Riyadh and May for Taif.ota) AM. ABOUAMMOH TABLE Mt ‘Surnmary statis foreach month and location Locations Month Saisie Dhahran Jeddah Khamis Madim Riyadh Tabouk Taf Mushait Jan, 99 ml 02ND OTR uo 17s S36 193535406710 242840882 Fob Max, 667 LOD 2 82 X89 798 Ika 88 701 491 s mal Sas 947 iss Mar Max, 28S 300 W970 996 Bay X 3525336272 2528 : a7) Om 4526S MSD NG 1407 Ape S18 3049 T9235 is 47723 as? D3 376 wos 2s KGL 82S wT May Max 23 mo 686 eo 060 x 0.66 26S 8 38a 5 Lis 4352S 60S BD Jun Max 00 003 22 oo oma ia oo 00 ser 0@ = 01 pa? : 00 00 52 167094 790 Jal Max, 00 200 8860 88 00a x oo 01062 og oy 28 9008s aE a Ave rn ne ma 02 120 00 00 2241 aor 5 00. 00 m6 as 379 ooh 2883 Sep Max. 00 lo ms 88 42 oo S84 x oo ous 398 oom 132 s oo 02 6a 287082 oon Ot May wt 400 sas ml eG x ole 106 zy ee tee LT : 392 120 T4627 6045 Now. Max sis 080 1eo 89010 aed 53200 Wada 037 2427 kat * n6s 29 ws6 TT Doe 82) 6AT Dec Max MO SSS SS 47 93 a Rss a4 5 pists? 65S 103DISHRIMGIHHON OF RAINFALL IN SAUISE ARABIA, 15}93 3. Rainfall Probabilities Probability plots of rainfall occurrence and rainfall amouras are very useful in choosing suitable probability models for fitting these variables. In arid regions rainfall occurrence can be studied more appropriately through the occurrence of dey spells or through the probability of no rain during specitic period. These probability plots ean be used to ‘compare the rainfall regimes of various sites Figure | illustrates the probability of the occurrence of dry monthy based on 21 cebservations for each month of the year for the sites. Altrough Dhahran has the maximum probability of having dry spells in the period June to September, Tabouk and, Jeddah also have a high probability of having dry spells for the same months of the year. Dhahran has the minimum probability of having dry spells in the period January to March, Riyadh is che second site that has a small probability of having dry spells for the same months of the year. There is 4 zero probability that May will be a dry month in Khamis Mushait. Khamis Mushait has minimum variation of probability values of dry spells and minimum averaged probability aver the 12 months. Figure 1 illustrates the probability of occurrence of a dry month in any year from 1966 to 1986. The maximum number of dry months (LI) are ip Tabouk (1973, 1978, and 1979) and Madina (1973), whereas the minimum number of dry months (2) are in Khamis JRFMAWS JASON OD Monthe Fig. 1. Probubiiy of cy noms for every month ar seven sits,94 (16 A.M ABOUAMMOH Mushait (1982), Madina (1978) and Taif (1986). Whereas the probability of having a dry month in any of the years 1966 through 1986 is illustrated in Figure 2, Figures 3 and4 represent the maximum and the mean amount of rainfall, respectively at the seven sites. In these figures, the axis (month) begins in July and ends in June for an easier comparison setween the sites. These figures show that Khamis Mushait hasa higher ‘mean rainfall than any other site in five months of the year, Madina has the lowest mean rainfall in four months ofthe year. Taif received maximum rainfall in seven months of the year, whereas Madina received the lowest maximum rainfall for four months of the year. Other comparisons with different sites can be based on these figures. 4 Fi 2 a Probability Distribution the time Let the rainfall occurrence follow a Poisson distribution with mean 1/4 interval between any Wo rainfall events has an exponential distribution with mean Thus, the number of rainfall events in a time interval is Poisson distributed with mean t/u. Rainfall amounts are assumed to be independent of their occurrences, mutually independent and exponentially distributed with mean 17. Such a model was used by Buishand (1977) for monthly totals. De Boer (1958) used a similar procedure to describe the distribution of rainfall months for monthly periods. The 66 68 70 72 74 76 78 80 a2 04 66 Years Fig. 2. Probability of dry months for cach year (1966-86),JASONDJFM AMS Monthe Fig. 3. Mean amount of monthly rafal at seven ste probability of receiving a rainfall of an amount less than or equal x during the time interval 0,0,i8 PIS x)= AIN=0)+ 2 ANS x| N=K) NHK), w where the number of rainfalls events N follows a Poisson distribution during the time interval with parameter 6¢ /p. Thus the probability distribution function is given by OF O04 se im Raysetter 3 & Ke dy, Q) rx) Spacois @ where A = p6 Integrating (2) over the total rainfall x gives PIN> I= | fay dx =e Jew mar er ALR =er 5 om96 18) AM AnOHAMMOH JASONDJFMAWM SG Months Fig 4 The maximum amount of rainfall t seven sites. ‘This implies vhat the number of rainfall during specific time interval is Poissin distributed. with a probability mass function PINK) =e" B/KI, k=O, 1, ‘The probability density function of precipitation totals for a given period is given by fey erven SOV, 4, e The random variable ¥ has a positive probability at x=0 with PX=0) = 6% “ Fisher and Cornish (1960) proposed this mixture of a gamma distribution for precipitation over arid regions. In fact Ozturk (1981) has used this model for fitting the monthly precipitation data of Whitestown meteorological station in Indiana, Abouam- ‘moh (1986) had found that the model used with probability density (3), is reasonable for fitting monthly data at Dhahran, Riyadh and Jeddah, Discussion of different estimation methods on fitting the monthly data in particular momest method, maximum likelihood. ‘method, empirical method and approximate maximum likelihood method have also been considered by the author.DISTRIBUTION OF RAINFALL IN SAUDI ARABIA, sor The simple empirical method is adopted for estimating the parameters of model (3). The ‘goodness of fit of the model is evaluated by the maximum absolute distance between the ‘empirical and the estimated distribution, Some properties of parameter estimates such as their means, variances, efficiencies relative to moment estimates and bias are discussed briefly in Ozturk (1984), Abouammoh (1986) found that estimating model (3) by using the empirical method leads to results similar to those results obtained by other methods of estimation. The probability density (1) can be expressed as foe ( oe )(A)ra Pax) 6) o)\y where J; stands for a modified Bessel function of order one, Fisher and Cornish (1960) have shown that the r-th cumulant of the distribution (2) is Ken, Thus the mean, the variance and the coefficient of skewness are B= O/A,0'= 2072, and == 3/426), respectively. ‘The probabilities of having no precipitation or some precipitation during a given time interval are q = PUX=0) = exp (-O)and p= 1-g = PUX>0) = |-exp-6), respectively. Also, the number of non-zero observations m, in a tandom sample of size n is a binomial random variable with parameters (n, p). The mean and the variance of mare np =r) and npg =ne*(I-e, respectively. The probability g=exp(-0)can be estimated fora randomsample»,..,x,,of which m are non-zero observations by g = I-(m/n). Hence 0 can be empirically estimated by 8 = Alog(1 ~ (m/n)), (6) ‘Therefore, asing the mean of the distribution F(x) implies X= O/- a Estimation of 8 and A given by relations (6) and (7) are easily calculated. The main drawoack of these estimates are 6 0, for m=0 and 6 for m=n, Table IV gives the estimated values for 6 and A in the seven sites, ‘5S. Concluding Remarks To test the goodness of ft one may evaluate the statistic D*, which is defined by Di = max, |Fx) - Fo, 1 Pla)is given by (5) whose parameters are estimated by the empirical method where n=:96 (20), AM. ANOUAMMON TABLE IW ‘of fle for seven meteorological stations Empirical estimation of param Monti Jan Feb Mar Apr May Jun Jui Aug Sep Oct Nov Dee Site Dhahran 2381 1.946 2351 12S od ~~ OSH O47 OAT K 0392 0260 0219 169 399 == omRs 0H” a.068 Jeklah 1099 0.84 0.184 0.37 27 = 009 = 0039 0.100 1.099 0.965 XK o0s6 0003 0079 01024 HO a02¢ = 1050 0025 0.085 9.112 Khamis 0965 647 1846 1986-1435 1658 1946 0.647 0560 0847 0.37 Masha —-K 061-0028 0.104 0.139 — 0377 0.136 0.169 O.118 0.095 0.063. 0.021 Madina 0687 0337 6.480 LSAT 0.742 0.184 O49 INL 0.184 0.272 O87 O67 K 007% 0.116 0.030 0.088 0.088 0.03% 0.008 0.447 0.612 0.067 0.069 0.134 Riyadh LABS LAIS GSK ROHS 1435 0.049 0.100 0.100 0049 0.154 0.560. 1658 K 0,174 0294 0.100 028260210 0028 1.623 0012 012 O18 0.129 11239 Tabok «D847 O47 0.965 0337 022 = = 00H = 0.279 0.965 0.560 0053 0.103 0.130 0029 9023 = = 02m = 08a aT 0.67 Tail a 0687 1438 Lesh 2351 OMT 0.560 0.742 L435 0.965 1435 0847 A (0.085 0.082 0.073 0.162 0.168 OLLIE OSE 0.274 0079 GEE 0.116 in Table IV and F,(x) is the empirical distribution function. Specifically FEL, Ay SHS Kyun 1. and xj, is the jth order statistic, =f, ...,.n. In fact there are not tables for evaluating the ‘goodness of fit for this mode! but smaller values for D* give some positive indication for better fitting. Table V contains the values of D* for every month of the year at the seven meteorological sites. The distribution of D* isnot available. Therefore, observed values of _D* can not be compared with its standard. Nonetheless, we can note that most of the observed values of D* are less than 0.2, Its also noted that relatively high values D* correspond to high maximum, or range, and high coefficient of variation. Specifically, values of D* for Marchat Dhahran (0.380) and January at Tabouk (0.302)are the highest valuesin Table 10, For other months at alsites, except where the model is not applicable, the proposed model fits the given data satisfactorily ‘The empirical method for estimating the parameters ofthe proposed modelis simple to carry out by using any calculator. The main drawback of this method is that itis not applicable for May at Khamis Mushait since it was a wet month during 1966 through 1986. For this month, other kown models for fitting precipitation data such as gamma, log normal and kappa types can be used, (Su2uki, 1980). Figures iffustrating the probability of dry months and maximum and mean amount of rainfall show a large variation in the distribution of monthly rainfalls at these sevenDISTRIBUTION OF RAINFALL IN SAUDE ARABIA, bape TABLEY The valve of D* statistics Site Dhahran Jeddah Khamis Madina Riyadh Tabouk = Taif Musa Moot Jan 0489 0237s gat 76 ons Feb 0255 008s Oo 0088 ons, Mar 0380 ome atts ok. 023s Apr on7 any alls as 8 ais May nom 109 : and? oes oa hs 5 : oor 00007 : 499 hu : ons 0.6018? 5 ous aa - - oss = OK? 087 ot) Sep - om = 00M = ous oe 062 ams ats ast oa Now Dos 07D 0.00% OLS Dec O13 0.9 atte tS sre sites. These plots are useful for quick ceference (o the rainfall regimes at the sites, ‘Thus ‘when planning for agricultural projects, vasiation in the rainfall behaviour has to be considered, ‘This is not an exhaustive study but it forms an additional data base for studying the distribution of monthly precipitation totals at the main meteorological sites in Saudi Arabia. Acknowledgement ‘The author is grateful to the referees for their usefull comments, References Abouammch, A.M. 1986, On Probability Disteibution of Monthly Precipitation Totals in Arid Regions Appl Sits. 2,863, Bosteuann, W. and Rothe. C: 1: 1976, "The Measurement of -mia Rain Rats from Weighing Rainags Recordings Appl Meteor 28, 161166 Buishand, T.A: 1977, Stockawie Modeling of oily Raimalt Sequences. Medd. Landbouwtogesch ‘Wageningse DeBoer, H-1: 1958,-On the Cumulative Proqueney Distribution of &-Day Period Amounts of Prespitation for any Station nthe Netherlands while k= 30, Ach, Met Geophys Bilin BB, 248-253, Drufocs,G.: 1977 "RadurDsrved Stat 1'Prcipitation Pater’ Ap Meteor 16 1020-1035, Fisher, R.A. and Cornish, EA. 1960, “The Percentile Points of Distributions Having Known Cumulans Teehnomerries 2, 20-225, Hamma, P1968, De Heart of Arabian Peninsula, (hn Arabic. Nae Riad. Kirkham, H: 1980 “Instantaneous Rainfall Rate, is Measurcment and is nflusnee on High-Voltage “Teansmisson Lines’. App. Metco. 19, 35-40 Linslay.J-RK., Kohler, M.A. and Paulhus. J. LH: 1982, Hydrology for Engincers, MeGiraw, Auckland109 22) A.M. ABOUAMMOH ‘Stern, RD. and Coe, Rs 1982, “The Use of Rainfall Models in Agricultural Pking’, Agri: Meteorol 24,3540, ‘Suruki, E 1980,"A Summarized Review of Theoretical Distibations Fitted to Climatic Factors and Markow ‘hain modlsof Weather Sequences, wth Some Exarpley in. hed al, ods), Statistical Climataigs Developmen Atmosphere Sciences 13, pp 120. Tayeb, FA. 1978, "The Role of the Ground Water in ligation and Drainage ofthe ALIasa Region af Eastezn Saudi Arabia", Unpublished M.A. Thesis, Eastem Michigan Universi. Ypsilanti Michigan, COnturk, A: 1981, On the Stay of a Probability Distbution of Precipitation Totals J. App. Meteor. 20, 1498150, (Ortuek, A 1984, “A Simple Method for Estimating the Parameters of Probability Distribution for Precipitation Total Proc. Contributed papers 44 Sers, ISI, pp. 4649UNEMODALITY OF DISCRETE DISTRIBUTIONS 1307 Theorem 2.1 : he necessary and sufficient (nés) condition for the discrete distribution b, to be uninodal with zero vertex is that the functions vee - a Py = BP ay and (2.2) Mey = Pre > 9 Pep pner are sone discrete distributions. (2.2) Proof: To prove the necessary part of the theorem, let P,, be a uninodal discrete distriution about zero. Thus, we need to show that the functions v, and W,_) , n eI are discrete distributions. From (2,1) by replacing n by m-1 one gets Ya = 2, fed = Pye ~ (OY Py (2.3) The difference of (2.1) and (2.3) gives Yy- Ve = Ue py + OF) py ay Consider the left hand side of the above equation to be sone number vy, that is ¥,=V,-V,q + thus, by taking the suma- ‘tion over n for all values of ne T, we get fv, = 1. Also, we notice that v,2 0 for all n ef, which yield that vy, ne T is a psm.f. and hence V, is a distribution function. In a simil- az manner one can alse show that W,_) ,n eI is a distribution function. Now, we prove the sufficient part of the theorem. If P is @iscrete distribution, ani v, - P, ete distribution, then one gets TAP, Ne Tis some discr- y, = AM, Py) (2.3) Since v2.0, Men p,2 Py for all n <0, andp, S pays for all n> 1. then n-l, the relation (2.5) can never give a clear relation betwen py and pj. Therefore the node of P,1348 ABOVAMMOK AND MASHHOUR is either at n= 0 orn= 1. By the sane argurent, condition (2.2) yields Way = UM (By > Pag which inplies that the distribution function P, is unimodal at either n= 0 or n=-1. Hence, by vertue of the conditions (2.1) and (2.2), the distribution ®, mist be unimodal about zero. This completes the proof. ‘The above theont can be generalized for any discrete distri- bution P, to be unimodal about a, where a ct, in the following corollary whose proof is omitted. Corollary 1: the discrete distribution P is unimodal about any point a¢ I, if and only if (iff) Yaen = Paen ~ P Paar, and Watnet ~ Pasi ~ ® Patned? are discrete distributions. he next result characterizes the unimodal discrete distrib- utions by sone differential equation and through their ch.fs.. nel Theorem ‘The ngs condition for the discrete distribution P., ‘hose ch.f. p(t), to be unimodal about zero is that the functions v(t) = pit) +i et) prit) wit) =e pry +i a - el) pricy are sone ch. fs. Proof: In the proof of theorem 2.1, we came to know that y= =n) pt im-1) py isa pam, iff P, is unimodal about zero. Then by multiplying both sides with o*? and taking the summation over all values of ne T, one gets