Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Standard view
Full view
of .
Look up keyword
Like this
0 of .
Results for:
No results containing your search query
P. 1


Ratings: (0)|Views: 2|Likes:
Published by SudharsananPRS

More info:

Categories:Types, Presentations
Published by: SudharsananPRS on Apr 03, 2014
Copyright:Traditional Copyright: All rights reserved


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less





Rainfall events prediction using rule-based fuzzy inference system
Somia A. Asklany
, Khaled Elhelow
, I.K. Youssef 
, M. Abd El-wahab
Cairo Regional Training Center, EMA, Cairo, Egypt 
Mathematics Department, Elmagmaa University, Elmagmaa, Saudi Arabia
Mathematics Department, Ain Shames University, Cairo, Egypt 
 Astronomy and Meteorology Department, Cairo University, Cairo, Egypt 
a r t i c l e i n f o a b s t r a c t
 Article history:
Received 22 September 2010Received in revised form 22 December 2010Accepted 24 February 2011
We are interested in rainfall events prediction by applying rule-based reasoning and fuzzylogic. Five parameters: relative humidity, total cloud cover, wind direction, temperature andsurface pressure are the input variables for our model, each has three membership functions.The data used is twenty years METAR data for Cairo airport station (HECA) [1972
1992] 30° 3
N, 31° 13
 E.and fiveyears METARdata for Mersa Matruh station (HEMM) 31°20
 N,27° 13
 E. Different models for each station were constructed depending on the availabledata sets. Among the overall 243 possibilities we have based our models on one hundredeighteen fuzzy IF
THEN rules and fuzzy reasoning. The output variable which has fourmembershipfunctions,takesvaluesfromzerotoonehundredcorrespondingtothepercentageforrainfall eventsgiven foreveryhourly data. Weusedtwo skillscorestoverify ourresults, theBrier scoreandthe Friction score.Theresultsarein highagreements with the recordeddata forthe stations with increasing in output values towards the real time rain events. Allimplementation are done with MATLAB 7.9.© 2011 Elsevier B.V. All rights reserved.
Rain forecastArti
cial intelligenceFuzzy logicFuzzy inference system
1. Introduction
The meteorological phenomenon chosen here to demon-strate fuzzy method is rainfall events. Information on rainforecasthasgreatimportanceinagriculturalareasspeciallyatthe northwest of Egypt where many crops irrigation mainlyontherainwater,alsoatCairoairportstationduetotheeffectof rain in visibility and in turn in the airport; so an expertsystem models based on fuzzy logic will be introduced toforecast rainfall events in Mersa Matruh (HEMM) station andCairo airport (HECA) station (Fig. 1). For the area underinvestigation statistical and synoptic studies show that theprobability for rain to be event increase in winter and springtime more than in autumn and occasionally happened insummer;becausetherainfallhappenedusuallywhenpassingthe fronts associated with the Mediterranean depressions.Rainfall forecastis an essentialand vitalprocessnowadays;each year hundreds died and were displaced by rains and
oods. Rainfall events can bepredicted using different models,Arnaud et al. (2007) built rainfall model based on the use of independent variables describing rainfall events and on theassumption that the process under study is a stationary one.Manel et al. (2009) investigated rainfall variability in southernTunisia by analyzing monthly and annual rainfall data. Melaniet al. (2010) use satellite images to study rainfall variabilityassociated with the summer African monsoon. Charabi and Al-Hatrushi (2010) study the observed relation between winterrainfall in Oman and the large-scale circulation and synopticactivity was examined on a monthly basis. Bookhagen (2010)presented a study for appearance of extreme monsoonalrainfall events in the Himalaya. However weather forecastingis one of the most imperative and demanding operationalresponsibilities carried out by meteorological services all overtheworld.Itisacomplicatedprocedurethatincludesnumerousspecialized
 Corresponding author. Tel.: +20 224501430; fax: +20 202 26849857.
E-mail address:
 somiaasklany@hotmail.com (S. Asklany).0169-8095/$
 see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.atmosres.2011.02.015
Contents lists available at ScienceDirect
Atmospheric Research
 journal homepage: www.elsevier.com/locate/atmos
to be taken in the uncertainty. Several authors (i.e. Brown andMurphy,1988;Elsnerand Tsonis,1992and manyothers)havediscussedtheuncertaintyassociatedwiththeweathersystems.Chaotic features associated with the atmospheric phenomenaalso have attracted the attention of the modern scientists(Sivakumar, 2000, 2001; Sivakumar et al., 1999; Men et al.,2004). Different scientists over the globe have developedstochastic weather models which are basically statisticalmodels that can be used as random number generatorswhose output resembles the weather data to which they havebeen
t (Wilks, 1999). Statistical models have the drawback that in most of the cases they depend upon several tacitassumptions regarding the system. But, a chaotic system likeatmosphere cannot be bound by any postulation. However,several studies have shown that NWP models and theirforecasts are subject to errors and biases because of thecomplex atmospheric uncertainties and the currently limitedknowledgeofthemathematicalformulationoftheatmosphericphysics and dynamics (Idowu and Rautenbach, 2009). Thenumerical models are based on the scheme of nonlinearoperator equations which prevail in the atmospheric system.But in the absence of any analog solution of this system of operator equations, numerical solutions based on differentassumptions are the only alternative. Furthermore, the chaoticbehaviors of these nonlinear equations sensitive to initialconditions make it more intricate to solve these equations(Guhathakurta, 2006). As a result,
awed forecast comes out.Arti
cial intelligence has been later investigated in weatherforecasting owing to their ability to deal with uncertainty,vagueness,incompleteandinexactspeci
cations,intuitionandqualitative information.The ability of fuzzy logic to handleimpreciseandinconsistentreal-worlddatamadeitsuitableforwide variety of application.Arti
cial intelligence (AI) is a term, in its broadest sense,meanstheabilityofamachineorartifacttoperformfunctionssimilar to those that characterize human thought. The termArti
cial Intelligence (AI) has been applied to computersystemsandprogramsthatcanperformtaskswhicharemorecomplex than straight forward programming, although stillfar from the realm of actual thought. AI consists of manybranches such as, Expert Systems (ESs), Arti
cial NeuralNetworks (ANNs), Genetic Algorithms (Gas) and Fuzzy Logic(FL) and various Hybrid Systems (HSs), which are combina-tions of two or more of the branches mentioned previously(Medske, 1996). AI technologies have a natural synergy thatcan be exploited to produce powerful computing systems. Athemethatcanbefoundinthesealternativesistheattempttomake up for de
ciencies in the conventional approaches. Insome cases, the goal is to produce better, more ef 
cient andeffective computing systems. AI has also been applied formodeling, identi
cation, optimization, prediction, forecast-ing, and control of complex systems (Mellit, 2008). But, thisrequires adding features associated with human intelligencesuch as learning and the ability to interpolate from currentknowledge. The appropriate use of intelligent technologiesleads to useful systems with improved performance or other
Fig. 1.
 Mersa Matruh and Cairo cities on the Egypt map.Source: http://www.maptown.com/geos/egypt.html"www.maptown.com/geos/egypt.html229
S. Asklany et al. /  Atmospheric Research 101 (2011) 228
characteristics that cannot be achieved through traditionalmethods (Rich and Knight, 1996).Fuzzy logic are a clever way to deal with vagueness thatoften faced in meteorological forecast
elds it can easilyincorporate expert knowledge into standard mathematicalmodels in the form of a fuzzy inference system (FIS). A FIS is anonlinear mapping of a given input vector to a scalar outputvector by using fuzzy logic. A FIS simulates the process of human reasoning by allowing the computer to behave lessprecisely than conventional computing. It is suitable forapproximate reasoning by using a collection of membershipfunctions and rules and is very powerful for modeling systemsthat are dif 
cult to represent by an accurate mathematicalmodel (Shu and Ouarda, 2008; Hansen, 2007). The
eld of meteorology commonly involves a system of concepts, princi-ples,andmethodsfordealingwithmodesofreasoningthatareapproximate rather than exact. The capability of dealing withimprecision gives fuzzy logic great potential for weatherforecast. In the recent years, fuzzy technique has drawnconsiderable attention towards handling this kind of complexand non-linear problems. The technique has been widelyapplied to many meteorological problems such as long termrainfallforecasting,Abrahametal.(2001),climateclassi
cation of atmosphericcirculation pattern, Bardossy et al. (1995), forecasting of temperature-humidity index using fuzzy logic approach byMitra et al. (2006) and fog forecasting using rule based fuzzyinference system Mitra et al. (2008).Rainfall forecast is one of the most complex elements insuch
elds since rainfall is a stochastic process, whoseupcoming event depends on some precursors from otherparameters such as the sea surface temperature (SST) formonthly to seasonal time scales, the surface pressure (SP) forweekly to longer than daily time scale and other atmosphericparameters for daily to hourly time scale. The latteratmospheric parameters could be temperature, relativehumidity and wind. Variability of weather and climaticfactors, especially those atmospheric parameters will be themajor forcing for precipitation event. If we could recognizesuch a variability pattern and use it for future trajectory,rainfall prediction is very much feasible. Wong et al. (2003)constructed fuzzy rule bases with the aid of back propagationneural networks and by the help of rule base developedpredictive model for rainfall. Shao (2000) used fuzzymembership functions for cloud amount, cloud type, windspeed and relative humidity to compose a fuzzy function of weather categorization for thermal mapping. Hansen (2003)presented a fuzzy case based prediction of cloud ceiling andits visibility. Hansen (2007) applied fuzzy k-nn weatherprediction system to improve the technique of persistenceclimatology by using past and present cases. Suwardi et al.(2006) used a neuro-fuzzy system for modeling wet seasontropical rainfall. Fallah-Ghalhary et al. (2009) has used fuzzyinferencesystemforpredictingrainfalltimeseriesdependingon the rainfall data corresponding to a certain period in thepast. Hasan et al. (2009) improved the primary fuzzy modelfor predicting rainfall by fuzzifying the wind speed, relativehumidity and temperature differences from the days beforerainfall data. In this study, we apply fuzzy logic and rule-based reasoning to predict rainfall events for every givenhourly data for two stations in Egypt Mersa Matruh (HEMM)and Cairo Air Port station (HECA) by generating membershipfunctions for each input parameter (relative humidity, totalcloud cover, wind direction temperature and surface pres-sure) and the output gives the percentage for rainfall event.
2. Data used in the models
The data used in building our models are twenty yearsMETAR  data for Cairo airport station [1972
1992] and
veyears Metar data for Mersa Matruh station [2003
rstwe separate the inputs parameters with their correspondingpresent weather data; separated set of data for trainingprocess. After building the models we selected the rainfallevent cases and six previous hours to study the behavior of the models.
3. Method
 3.1. Fuzzy logic 
Our attempt is to forecast rainfall with the help of fuzzylogic based approximate reasoning. This process uses theconceptofapurefuzzylogicsystemwherethefuzzyrulebaseconsists of a collection of fuzzy IF
THEN rules. The fuzzyinferenceengineusesthesefuzzyIF
THENrulestodeterminea mapping from fuzzy sets in the input universe of discourseto fuzzy sets in the output universe of discourse based onfuzzylogicprinciples.Inordertobuildourmodelswede
nedthe fuzzy sets consist of 
ve parameters: relative humidity,total cloud cover, wind direction, temperature and surfacepressureare the inputvariables for our model;each hasthreemembership functions with single output which is rain eventpercentage.Under the fuzzy set theory, elements of a fuzzy set aremappedtoauniverseofmembershipvaluesusingafunction-theoretic form belonging to the closed interval from 0 to 1(Mellit, 2008).FL is very useful in modeling complex and imprecisesystems, and fuzzy set theory is a powerful tool and itsapplications have rapidly increased with establishing itsutility in numerous areas of the scienti
c world. Any systemconsisting of vague and ambiguous input variables maycontribute to an ultimate effect. The fuzzy logic possibilityand its degree of effect due to the ambiguous input variablesare considered by some as being generated in the humanmind and is often referred to as expert knowledge. Thisexpert knowledge is the accumulation of knowledge andideas as a result of the expert's experience in a particularsystem;hence,decision-makingprocessesmaybeconsideredas fuzzy expressions perceived by the expert (Hasan et al.,2009).The fuzzy theory
rst proposed by L.A. Zadeh (1965),operated through three main steps.1.
rst step is to determine the de
nitiondomain of each variable based on the ranges of input andoutput variables in actual conditions.2.
 Fuzzy rules determination and fuzzy inference:
 Based on theexperience and knowledge of experts, the language rulesof determination were transferred into the executablefuzzy syntax for inference.
 S. Asklany et al.  Atmospheric Research 101 (2011) 228

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->