A low cost acous:c disdrometer as cizen sensor for rainfall … · 2016. 10. 10. · Andrea...
1
Andrea Antonini (1), Alessandro Mazza (1,2), Samantha Melani (1,2), Francesco Saba:ni (2) and Alberto Ortolani (1,2) Affilia:ons: (1) LaMMA Consor:um (2) CNR-IBIMET mail to: [email protected] web: www.lamma.rete.toscana.it PRELIMINARY RESULTS A low cost acous:c disdrometer as ci:zen sensor for rainfall measurements INTRODUCTION In the last decades the increasing occurrence of extreme weather events in urban areas has highlighted some weakness and gaps of the observing network systems used by civil protec:on organiza:ons. More specifically, the precipita:on parameter is characterised by very high spa:al and temporal variability so that the measurement networks fail to monitor it with the necessary precision to provide meaningful informa:on at urban scale. Furthermore, the measurement networks are rarely able to dis:nguish the type of precipita:on with resolu:ons that significantly can impact on the urban areas. Polarimetric weather radars are instruments that can extract informa:on about the type of precipita:on, but they are expensive and oBen characterized by difficul:es during installa:on procedures (obtaining authoriza:ons, risk for popula:on exposure to electromagne:c fields, etc.). Disdrometers are able to characterise the precipita:on type also on the different height levels of the ver:cal atmospheric profile above their installa:on sites. The second processing algorithm is in the frequency domain: Short Time Fourier Transform (STFT) was performed. The preliminary STFT analysis in the frequency domain shows very clearly the impact of the hydrometeors, as the signal consists of a wide spectral dynamic. The selected case study, rela:ve to a short :me period in which eight hydrometeors impact the sensor, shows that the eight drops can be clearly recognized in the frequency domain but not easily in the :me domain. This behaviour persists even in the quan:ta:ve analysis of the signal, as the frequency domain is more sensi:ve and can more directly extract informa:on. For this case, a short :me interval (1 s) has been chosen to avoid too many impacts of hydrometeors, a circumstance that would make unclear the example. 8th IPWG and 5th IWSSM Joint Workshop Bologna, 3-7 October, 2016 The instrument is an acous:c disdrometer working with the same principle as an ordinary microphone that converts sound waves in electric signals. A piezo detector, fixed to the boNom layer of the acous:c disdrometer, converts the vibra:ons caused by the drops impac:ng the instrument to an electric signal, which is the input for data logger and processing. The disdrometer (see Figure) was realised by assembling: a sheet of glass mounted on a wood frame; a piezoelectric element (used as a contact microphone) fixed to the boNom layer of the glass; a bipolar cable connected to the bipolar output of the sensor. In the proposed prototype implementa:on, the sound card of a personal computer was used as the signal acquisi:on system. The limited sampling frequency of the sound card (44kHz) is sufficient to extract the needed spectral and amplitude informa:on :ed to precipita:on occurrence. THE ACOUSTIC DISDROMETER THE METHODOLOGY The signal processing is composed of two analysis algorithms; the first one is in the :me domain, where the signal amplitude can be easily managed and the waveforms caused by the raindrop are characterised by an ini:al peak followed by damping oscilla:ons due to the relaxa:on of the piezo-element to the equilibrium state. The hydrometeors impac:ng the sensor are dis:nguishable by displaying their amplitude versus :me, as the amplitude of the ini:al peak is much greater than the background noise. The peak amplitude is correlated to the size of the impac:ng drop, as well as the transient :me. The number of raindrops falling in a certain :me interval is correlated to the rain rate in the same :me interval. The chosen :me interval is 1 minute. Some analysis limita:ons are due to the background noise amplitude introduced by the cable and the sensor and the external acous:c noise (i.e., air condi:oning machines in proximity of the sensors, traffic noises, airplanes, etc.) SENSITIVITY TESTS Some laboratory tests were conducted in order to inves:gate acous:c disdrometer sensi:vity to raindrop diameter. Raindrops of 1, 3 and 5 mm diameters were falling down from a height of 18m (the roof of a building) in order to reach their terminal velocity before the impact with the instrument, thus simula:ng a real rainfall drop. Raindrop diameter [mm] TIME N° of drops Mean dura:on [s] Normalized Mean amplitude 1 16:05:35 13 0,017 0,13 3 16:01:11 56 0,039 0,43 5 15:44:52 27 0,046 0,76 The results show clearly the capability of the system in dis:nguishing different drop sizes. The algorithm implemented for drop analysis provides as output the drop dura:on and maximum amplitude, together with the number of detected drops in a :me frame of 55s. In the Table, the signal due to the various drop diameters differs significantly for dura:on and amplitude. 1mm 3mm 5mm CONCLUSIONS and FUTURE WORKS The preliminary results are encouraging versus a reliable use of this low cost instrument; its sensi:vity to drop diameter is promising in detec:ng different precipita:on regimes. Very preliminary comparison with raingauge (:pping bucket resolu:on 0.2mm sampling :me 1m) have been made for calibra:on purposes. An extended dataset with various rainrate intensi:es is needed for obtaining a calibra:on rela:onship. Future developments are aimed to refine the algorithms using different piezoelectric sensors in different sites under various precipita:on regimes. Finally, also hail par:cles will be inves:gated. Some sta:s:cal analyses during the first two months of 2016 were performed in order to verify the system rain rate es:ma:on capability, using a :pping bucket raingauge as reference. The figures show the histograms of the number of detected drops, drop signal dura:on and normalized amplitude for a rainrate of 1.2 mm/h (a :pping bucket of 0.2mm in 1 minute). The histograms can provide informa:on about most recurrent values to be used for calibra:on purposes. As further parameter the sum of absolute values of samples is provided.
Transcript of A low cost acous:c disdrometer as cizen sensor for rainfall … · 2016. 10. 10. · Andrea...
AndreaAntonini(1),AlessandroMazza(1,2),SamanthaMelani(1,2),FrancescoSaba:ni(2)andAlbertoOrtolani(1,2)Affilia:ons:(1)LaMMAConsor:um(2)CNR-IBIMET
mailto:[email protected]:www.lamma.rete.toscana.it
PRELIMINARYRESULTS
Alowcostacous:cdisdrometerasci:zensensorforrainfallmeasurements
INTRODUCTIONInthelastdecadestheincreasingoccurrenceofextremeweathereventsinurbanareashashighlightedsomeweaknessandgapsoftheobservingnetworksystemsusedbycivilprotec:onorganiza:ons.Morespecifically,theprecipita:onparameterischaracterisedbyveryhighspa:alandtemporalvariabilitysothatthemeasurementnetworksfailtomonitoritwiththenecessaryprecisiontoprovidemeaningfulinforma:onaturbanscale.Furthermore, the measurement networks are rarely able to dis:nguish the type of precipita:on with resolu:ons that significantly can impact on the urban areas.Polarimetricweather radars are instruments that canextract informa:onabout the typeofprecipita:on,but theyareexpensiveandoBen characterizedbydifficul:esduringinstalla:onprocedures(obtainingauthoriza:ons,riskforpopula:onexposuretoelectromagne:cfields,etc.).Disdrometersareabletocharacterisetheprecipita:ontypealsoonthedifferentheightlevelsofthever:calatmosphericprofileabovetheirinstalla:onsites.
The second processing algorithm is in the frequency domain: Short Time Fourier Transform (STFT) wasperformed. The preliminary STFT analysis in the frequency domain shows very clearly the impact of thehydrometeors,asthesignalconsistsofawidespectraldynamic.Theselectedcasestudy,rela:vetoashort:me period in which eight hydrometeors impact the sensor, shows that the eight drops can be clearlyrecognized in the frequencydomainbutnoteasily in the:medomain. Thisbehaviourpersists even in thequan:ta:veanalysisof thesignal,as the frequencydomain ismoresensi:veandcanmoredirectlyextractinforma:on. For this case, a short :me interval (1 s) has been chosen to avoid too many impacts ofhydrometeors,acircumstancethatwouldmakeuncleartheexample.
8thIPWGand5thIWSSMJointWorkshopBologna,3-7October,2016
Theinstrumentisanacous:cdisdrometerworkingwiththesameprincipleasanordinarymicrophonethatconvertssoundwavesinelectricsignals.Apiezodetector,fixedtotheboNomlayeroftheacous:cdisdrometer,convertsthevibra:onscausedbythedropsimpac:ngtheinstrumenttoanelectricsignal,whichistheinputfordataloggerandprocessing.Thedisdrometer (seeFigure)was realisedbyassembling:asheetofglassmountedonawood frame;apiezoelectricelement(usedasacontactmicrophone)fixed to theboNom layerof theglass;abipolarcableconnected to thebipolaroutputof thesensor. In theproposedprototype implementa:on, thesoundcardofapersonalcomputerwasusedas thesignalacquisi:onsystem. The limited sampling frequency of the sound card (44kHz) is sufficient to extract the needed spectral and amplitudeinforma:on:edtoprecipita:onoccurrence.
THEACOUSTICDISDROMETER
THEMETHODOLOGYThesignalprocessingiscomposedoftwoanalysisalgorithms;thefirstoneisinthe:medomain,wherethesignalamplitudecanbeeasilymanagedandthewaveformscausedbytheraindroparecharacterisedbyanini:al peak followedby dampingoscilla:ons due to the relaxa:onof thepiezo-element to the equilibriumstate.Thehydrometeorsimpac:ngthesensoraredis:nguishablebydisplayingtheiramplitudeversus:me,astheamplitudeoftheini:alpeakismuchgreaterthanthebackgroundnoise.Thepeakamplitudeiscorrelatedtothesizeofthe impac:ngdrop,aswellasthetransient:me.Thenumberofraindropsfalling inacertain:me interval is correlated to the rain rate in the same:me interval. The chosen:me interval is 1minute.Someanalysislimita:onsareduetothebackgroundnoiseamplitudeintroducedbythecableandthesensorand the external acous:c noise (i.e., air condi:oning machines in proximity of the sensors, traffic noises,airplanes,etc.)
SENSITIVITYTESTSSome laboratory tests were conducted inorder to inves:gate acous:c disdrometersensi:vity to raindrop diameter. Raindrops of1, 3 and 5 mm diameters were falling downfromaheightof18m(theroofofabuilding)inorder to reach their terminal velocity beforethe impact with the instrument, thussimula:ngarealrainfalldrop.
Raindropdiameter[mm]
TIME N°ofdrops Meandura:on
[s]
NormalizedMean
amplitude
1 16:05:35 13 0,017 0,133 16:01:11 56 0,039 0,435 15:44:52 27 0,046 0,76
The results show clearly the capability of thesystemindis:nguishingdifferentdropsizes.
The algorithm implemented for drop analysisprovides as output the drop dura:on andmaximumamplitude,togetherwiththenumberofdetecteddropsina:meframeof55s.
IntheTable,thesignalduetothevariousdropdiametersdifferssignificantlyfordura:onandamplitude.
1mm
3mm
5mm
CONCLUSIONSandFUTUREWORKSThepreliminaryresultsareencouragingversusareliableuseofthislowcostinstrument;itssensi:vitytodropdiameterispromisingindetec:ngdifferentprecipita:onregimes.Verypreliminarycomparisonwithraingauge(:ppingbucketresolu:on0.2mmsampling:me1m)havebeenmadeforcalibra:onpurposes.Anextendeddatasetwithvariousrainrateintensi:esisneededforobtainingacalibra:onrela:onship.Futuredevelopmentsareaimedtorefinethealgorithmsusingdifferentpiezoelectricsensorsindifferentsitesundervariousprecipita:onregimes.Finally,alsohailpar:cleswillbeinves:gated.
Somesta:s:calanalysesduringthefirsttwomonthsof 2016 were performed in order to verify thesystemrainratees:ma:oncapability,usinga:ppingbucketraingaugeasreference.The figures show the histograms of the number ofdetecteddrops,dropsignaldura:onandnormalizedamplitude for a rainrate of 1.2 mm/h (a :ppingbucketof0.2mmin1minute).Thehistogramscanprovideinforma:onaboutmostrecurrentvaluestobeusedforcalibra:onpurposes.
Asfurtherparameterthe sum of absolutevalues of samples isprovided.
