Transcript
Int.J. Electron.Commun.(AEÜ) 68 (2014) 33–36 ContentslistsavailableatScienceDirect International Journal of Electronics andCommunications (AEÜ) jo urnalhomepage:www.elsevier.com/locate/aeue Time series prediction of rain attenuation from rain ratemeasurement using synthetic storm technique for atropical location Dalia Das a ,Animesh Maitra b , ∗ a DepartmentofElectronicsandTelecommunicationEngineering,MeghnadSahaInstituteof Technology,TechnoComplex,Madurdaha,Kolkata700150,India b S.K.MitraCentreforResearchinSpaceEnvironment,InstituteofRadioPhysicsandElectronics,UniversityofCalcutta,Kolkata700009,India a r tic le inf o Articlehistory: Received16December2012Accepted15July2013 Keywords: RainattenuationSyntheticstormtechniqueTimeseriespredictionTropicallocation ab st ra ct A comparisonof measured attenuation series with the attenuationseries obtained from rainrate mea-surement by usingsyntheticstorm techniqueismade for Ku band signalatatropicallocation.Validityof the model istestedforthe long-termstatistics interms of the cumulativedistribution of attenuationoccurrence and fade duration. Applicabilityof the model isalso shownto be valid event-wise. It hasbeendemonstrated that the long term statistics of predicted rainattenuation are insensitive tostormtransla-tion speed. No significant differencesare foundwhencumulative distributions of predicted attenuationvalues are compared for different data samplingintervals. It hasbeen observed that there exists agoodcorrelationbetweenthepredictedand measured values of attenuationfor at least80%of the events. © 2013 Elsevier GmbH. All rights reserved. 1.Introduction Frequenciesabove10GHzareof primaryinterestinsatellitecommunicationsystems,sincetheyprovidelargertransmissionbandwidthandhigherdatarate.However,theuseof thesefre-quencybandsislimitedbydifferentpropagationeffectmainlyduetorainattenuation.If timeseriespredictionofrainattenuationispossible,fadecountermeasuretechniquessuchasadaptivecontrolofsignalpower,codinganddataratecanbeeffectivelyimple-mented.Themethodoftimeseriespredictionforrainattenuationhasbeenpresentedin[1]. Experimentaldataforrainattenuationtodevelopchannelmodelarenotalwaysavailableandoftentheyexistonlyforspecificsites,frequenciesandelevation.Butalargeset ofrainratedatais availableworldwide.Asrainattenuationisstronglycorrelatedwithrainrateintensity,timeseriespredictorofrainratecanbe easilyconvertedintorainattenuationpredictorbyusingso-calledsyntheticstormtechnique(SST).SSThasbeenproposedin[2]toconvertinstantaneousrainrateintoattenua- tionundersomeassumption.Sofar,validityoftheSSTmodelispresentedintermsofyearlycumulativedistribution[2–5]. In[6,7]validationresultsarepresentedonaneventbyeventbasis,butonlyeventdurationandpeakattenuationarecomparedforV bandsignalsfortemperateregion.Inthispaper,measuredrainrateseriesduringa raineventisconvertedintoattenuationseriesfortheKubandsignalfora tropi- ∗ Correspondingauthor.Tel.:+919433733756;fax:+91 3323515828. E-mail address:
[email protected](A.Maitra). calregion.Timeseriespredictionofattenuationis doneduringraineventsusingthemethoddescribedin[1]. However,inthepresentcase,SSTconvertedattenuationvaluesareconsideredasinputsinsteadof actualattenuationmeasurements.Validityofthesyn-theticstormtechniqueis notonlytestedevent-wisebutalsowithlongtermstatistics.Resemblancebetweenmeasuredandpredictedeventisalsoshownbycalculatingcrosscorrelationcoefficient.Stormtranslationspeedsuitableforourregionis alsoselectedfromexperimentalresults. 2.Experimentaldata Propagationmeasurementsoveranearth-spacepathhavebeencarriedoutatKolkata,India(22 ◦ 34 ′ N,88 ◦ 29 ′ E),a tropicallocationbyreceivinga Kubandsignalatfrequency11.172GHztransmittedwithhorizontalpolarizationfromsatelliteNSS-6(geostationaryat95 ◦ E)at anelevationof63 ◦ , sinceJune2004[8]. Thereceivedsig-nalisdownconvertedtoanL-bandfrequencybythelownoiseblockconverter(LNBC)andfedtothespectrumanalyzerthatisusedasthereceiverformonitoringthesatellitesignallevel.Thesignallevelmeasurementsarerecordedwithadataloggerandstoredin a PC.Further,therainfallratesatthesatellitereceiversitehavebeenmeasuredsimultaneouslybyanopticalraingauge(ORG).Thedynamicrangesforrainrateandattenuationmeasurementsare500mm/h and20dBrespectively.Theminimumdetectablechangeinrainfallrateis0.2 mm/h andrainattenuationis0.1dB.Therecordedrainrateandattenuationdataarepassedthrougharaisedsquarecosinefilterwithcutofffrequency0.025Hztoeliminatethescintillationeffectsandotherfastfluctuations.Inthepresentstudy, 1434-8411/$–seefrontmatter © 2013 Elsevier GmbH. All rights reserved. http://dx.doi.org/10.1016/j.aeue.2013.07.008 34 D.Das,A. Maitra/ Int.J.Electron.Commun.(AEÜ) 68 (2014) 33–36 Fig.1. (a)ComparisonbetweenthemeasuredattenuationvalueswiththepredictedvaluesobtainedfromSSTfortheraineventof 17thJune,2007.Thestormspeedistakenas v =8m/s and(b)comparisonbetweenthemeasuredattenuationvalueswiththe predictedvaluesobtainedbytimeseriespredictorfortheraineventof17thJune,2007.Measurementneededbythepredictoris takenasthe attenuationvaluesobtainedbySST. thefouryearmeasurementperiod(2005–2008)hasbeenconsid-eredduringwhichtotal694raineventsareobserved.Inthispaper,whencumulativedistributionis calculated,theentiretimespanof measurementisconsidered. 3.Modeltestingwithexperimentaldata Thesyntheticstormtechnique(SST)convertsarainratetimeseriesrecordedata givenlocationintoa signalattenuationtimeseries.Thisconversionrequirestheknowledgeaboutthelengthof thesignalpaththroughtheraincell,thevelocity( v )of theraincellandtherainrate( R )atthelocationunderinvestigation.Thephysi-calandmathematicalfundamentalsof themethodaredescribedin[2].Theverticalstructureof theprecipitationmediumismodelledwithtwolayers[2],layerAwithraindropsat20 ◦ CandlayerBwithmeltinghydrometeorsat0 ◦ C.TheinputparametersneededbytheSSTmodelforourregionareconsideredasfollows.Thealtitudeabovesealeveloftheearthstationis H S = 0.025km.Accordingto[9]theheightof theprecipitation(rainandmelt-inglayer)abovesealevelusedin thesimulationis calculatedas H B =5km. Also,thedepthof themeltinglayer( h )isconsideredtobe0.4kmregardlessofthelatitude.Accordingto [2]theheightabovesealevel, H A , oftheupperlimitof layerAisgivenby: H A = H B − h = 4 . 6kmTheradiopathlengthsaregivenby L A = H A − H S sin( ) = 5 . 5836km L B = H B − H S sin( ) = 5 . 135kmTheparameters k and ˛ necessaryto relaterainfallratetothespecificrainattenuation(dB/km)arecalculatedfrom[10]. We have Fig.2. Comparisonbetweenthecumulativedistributionsof predictionerrors(%)due toSSTpredictionasshowninFig.1(a),andtimeseriespredictionwithSST valuesasinputasshowninFig.1(b). useddifferentstormspeeds v = 1–12m/s toshowthesensitivityof thisparametertotheSSTmodel.Themeasuredrainratevaluesfortheraineventon17June2007areconvertedtoattenuationvaluesusingSSTandcomparedwiththeactualmeasurementsinFig. 1(a).Goodmatchinghasbeenobservedbetweenmeasurementandprediction.TheseSSTpredictedvaluesarenowusedasmea-surementsforthemethoddescribedin [1]to predictthetimeseriesofrainattenuationfortheraineventon17June2007.ThetimeseriespredictedvaluesarecomparedwiththeactualattenuationmeasurementsinFig.1(b).Fig.2givesthecomparisonbetween thecumulativedistributionsof thepredictionerrors(%)occurredinFig.1(a)and(b).Forthefirstcase,shownin Fig.1(a),theerroroccurredonlyduetoSST prediction.Whereasinthesecondcase,showninFig.1(b),theerroris duetobothSSTpredictionandtimeseriesprediction,resultingin asmallincreaseinthetotalerror.However,theoverallerroris stillsmallindicatingthattheSSTpre-dictedvaluescanbeconsideredastheinputto thetimeseriespredictorin theabsenceofactualattenuationmeasurements. Fig.3. ComparisonbetweenthemeasuredattenuationvalueswiththepredictedvaluesobtainedfromSSTfortheraineventof17thJune,2007with(a) 10ssamplingintervaland (b)60ssamplinginterval. D.Das,A. Maitra/ Int.J. Electron.Commun.(AEÜ) 68 (2014) 33–36 35 Ifweusedifferenttime resolutionforrainratemeasurements,asignificantdifferenceisfoundforhigherattenuationvaluesinasingleraineventasisevidentfromFig.3.Althoughoverallaccuracy ofattenuationpredictionisbetterforsmallersamplingintervalasexpectedwithSST,forhigherintervalbettermatchingis observedforhigherattenuationvalues.ThisisbecauseofthefactthatforsmallersamplingintervalhighrainratevaluesatthereceivingsitearerecordedmorefrequentlygivinghighSSTestimatesofatten-uationvalueswhichdo notmatchwithmeasuredattenuationsashighrainratesmay not occurovertheentiresignalpath.Sothematchingbetweenpredictionandmeasurementforhigherattenu-ationvaluesisshowntobeapparentlypoorerforsmallersamplinginterval.Thecumulativedistributionsoftotalsignalattenuationforboththemeasuredandthesynthesizedeventsforthecompleteperiod(averagedoverthewholefouryearmeasurementperiod)areshowninFig. 4. Theagreementbetweenboththedistributionsisquitegood.In thispaper,whencumulativedistributionis cal-culated,theentiretimespanof measurementis considered.FadedurationstatisticscanalsobepredictedbytheSST.Comparisonsbetweenthepredictedandmeasuredcumulativedistributionsof fadedurationsfordifferentthresholdsareshowninFig.5.Agoodmatchinghasbeenobservedbetweenthemeasuredandpredictedstatistics.Althoughforsingleraineventtimeseriesisdifferentfordifferentsamplinginterval(Fig.3),whencumulative Fig.4. Comparisonbetweenthemeasuredandpredictedrainoccurrencestatisticsfortheperiod2005–2008. distributionsof rainattenuationresultingfromSSTsimulationforsamplingintervalnamely10, 30and60sarecompared,nosignificantdifferencesarefoundasindicatedinFig.6(a).In Fig.6(b),cumulativedistributionsof predictedrainattenuationvaluesare plottedfordifferentstormtranslationspeedalongwiththemeasurement.FromFig.6(b)itisclearthatlongterm statisticsderivedfromtheSSTmodelis almostinsensitivetostorm Fig.5. Comparisonbetweenthemeasuredandpredictedfadeduration‘statisticsfortheperiod2005–2008fordifferentthreshold. Fig.6. Cumulativedistributionsofpredictedsignalattenuationvaluesfor(a) differentsamplingtimeintervaland(b) differentstormspeed. 36 D.Das,A. Maitra/ Int.J.Electron.Commun.(AEÜ) 68 (2014) 33–36 Fig.7. Comparisonbetweenthemeasuredattenuationvalueswiththe predictedvaluesobtainedfromSSTfortheraineventof17thJune, 2007withstormspeed(a)8m/sand (b)12m/s. Fig.8. Cumulativedistributionsof crosscorrelationcoefficientsbetweenmeasuredand predictedattenuationvaluesforalltheraineventsofthefouryearperiod,2005–2008. speed.ForbothFig.6(a)and(b)goodmatchinghasbeenobserved betweenthestatisticsof measuredattenuationvaluesandallthepredictedvalues.Fig.7showshowthevalueof thestormspeed,usedin thepredictions,affectsthetimeseries.Forhighervaluesof thestormspeed,thepeakattenuationbecomeslargeratagivenrainrate.Thezerolagcrosscorrelationcoefficient oftheindividualeventsforthefouryearperiod2005–2008hasbeencalculatedfordifferentwindspeed v =1–12m/s.Fig.8showsthecumulativedistributions ofthecomputed withdifferentstormspeed.FromFig. 8itisclearthatifwechoose v = 8m/s,cumulativedistributionsfor showhighervaluesindicatingthatmatchingbetweenpredictionandmeasurementisgood.Butifweincrease v above8m/s ordecreasebelow8m/s,cumu-lativedistributionsfor showlowervalues.Thisindicatesthatforourregion,suitablestormtranslationspeedis8m/s. 4.Conclusion Developmentof channelmodelto predicttimeseriesofrainattenuationisnotalwayspossibleindifferentclimaticareasandatdifferentfrequencybandsduetolackofattenuationmeasure-ments.However,rainrecordingsareeasiertoobtain.Thesecanbeconverteddirectlyintorainattenuationseriesbyusingsyn-theticstormtechnique.In thispaper,validityoftheSST modelhasbeenpresentedforKu-bandsignalforatropicallocation,India.Themodelisvalidatednot onlyona cumulativedistributionbasis,butalsoonaneventbyeventbasis.Thecumulativedistributionofsignalattenuationandfadedurationstatisticsfrompredictedvaluesmatchedwellwiththatobtainedfrommeasuredvalues.SSTdoesnotsignificantlydependonthesamplingrateatwhichrainrecordingsaretaken.Long-termstatisticsofSSTsimulatedresultisinsensitiveto stormspeed.Butforindividualevents,8m/sstormspeedgivesbestresultforIndianregion.ItcanbeconcludedthattheSSTmodelcanbeusedforpredictingtimeseriesof sig-nalattenuationata tropicalregiontoimplementfademitigationtechnique. Acknowledgments Thisworkhasbeensupportedbythegrantsundertheprojectentitled“Integratedstudiesonwatervapour,liquidwatercontentandrainoftropicalatmosphereandtheireffectsonradioenvi-ronment”,fundedbyIndianSpaceResearchOrganization(ISRO),Bangalore,India,beingimplementedatS.K.MitraCentreforResearchin SpaceEnvironment,Universityof Calcutta. References [1]DasD,MaitraA.Timeseriespredictorof Ku–bandrainattenuationoveran earth–spacepathatatropicallocation.IntJ SatellCommunNetw2012;30(January/February):19–28.[2]MatriccianiE.Physical–mathematicalmodelofthedynamicsofrainatten- uation basedonrainratetimeseriesandatwo-layerverticalstructureof precipitation.RadioSci1996;31(March–April(2)):281–95. [3]MatriccianiE.Predictionof fadedurationsdue toraininsatellitecommunica-tionsystems.RadioSci1997;32(March–April(3)):935–41. [4]MatriccianiE,RivaC.ThesearchforthemostreliablelongtermrainattenuationCDFofaslantpathandtheimpactonpredictionmodels.IEEETransAntennasPropag2005;53(September(9)):3075–9.[5]KanellopoulosSA,PanagopoulosAD,MatriccianiE,KanellopoulosJD.Annualand diurnalslantpathrainattenuationstatisticsinAthensobtainedwiththe syntheticstormtechnique.IEEETransAntennasPropag2006;54(August(8)):2357–63.[6]FontanFP,NunezA,ValcarceA,FiebigUC.Convertingsimulatedrain-rateseriesintoattenutionseriesusingthesyntheticstormtechnique.In:COST280PM91043rdinternationalworkshop.2005.[7]Sánchez-LagoI,FontánFP,Mari˜noP,FiebigUC.Validationof thesyntheticstormtechniqueaspartof a time-seriesgeneratorforsatellitelinks.IEEEAntennasWirelessPropagLett2007;6:372–5.[8]MaitraA,ChakravartyK, BhattacharyaS,BagchiS. PropagationstudiesatKu-bandoveranearth-spacepathatKolkata.IndJ RadioSpacePhys2007;36:363–8.[9]Rainheightmodelforpredictionmethods,ITU-RRecommendations,Propaga-tioninNonionizedMedia,Rec.839,Geneva;1992.[10]MaggioriDD.Computedtransmissionthroughraininthe1–400GHzfre- quencyrangeforsphericalandellipticaldropsandanypolarization.AltaFreq1981;50:262–73.
