Nighttime wind and scalar variability within and above an Amazonian canopy

17 Pages·2017·2.58 MB·English

Atmos.Chem.Phys.,18,3083–3099,2018
https://doi.org/10.5194/acp-18-3083-2018
©Author(s)2018.Thisworkisdistributedunder
theCreativeCommonsAttribution4.0License.
Nighttime wind and scalar variability within and
above an Amazonian canopy
PabloE.S.Oliveira1,OtávioC.Acevedo1,MatthiasSörgel2,AnywhereTsokankunku2,StefanWolff2,
AlessandroC.Araújo3,RodrigoA.F.Souza4,MartaO.Sá5,AntônioO.Manzi5,andMeinratO.Andreae2
1DepartamentodeFísica,UniversidadeFederaldeSantaMaria,Av.Roraima1000,SantaMaria,RS,Brazil
2BiogeochemistryDepartment,MaxPlanckInstituteforChemistry,P.O.Box3060,55020Mainz,Germany
3EmpresaBrasileiradePesquisaAgropecuária(EMBRAPA),Trav.Dr.EnéasPinheiro,Belém,PA,Brazil
4UniversidadedoEstadodoAmazonas(UEA),Av.DarcyVargas1200,Manaus,AM,Brazil
5InstitutoNacionaldePesquisasdaAmazônia(INPA),Av.AndréAraújo2936,Manaus,AM,Brazil
Correspondence:PabloE.S.Oliveira([email protected])
Received:6July2017–Discussionstarted:10July2017
Revised:5December2017–Accepted:12December2017–Published:5March2018
Abstract. Nocturnal turbulent kinetic energy (TKE) and 1 Introduction
fluxesofenergy,CO andO betweentheAmazonforestand
2 3
the atmosphere are evaluated for a 20-day campaign at the The turbulence structure above forested canopies has been
Amazon Tall Tower Observatory (ATTO) site. The distinc- animportantsubjectofresearchoverthepastdecades.Such
tionofthesequantitiesbetweenfullyturbulent(weaklysta- knowledgeisessentialtoanswerrelevantscientificquestions
ble)andintermittent(verystable)nightsisdiscussed.Spec- suchasthequantificationoftheexchangeofscalarsbetween
tralanalysisindicatesthatlow-frequency,nonturbulentfluc- forestedecosystemsandtheatmosphere.Someoftheprecur-
tuationsareresponsibleforalargeportionofthevariability sorstudiesinthisfieldhavebeenperformedintheAmazon
observedonintermittentnights.Intheseconditions,thelow- region during projects such as the Atmospheric Boundary-
frequency exchange may dominate over the turbulent trans- LayerExperiment(ABLE2Aand2B;Fitzjarraldetal.,1988;
fer. In particular, we show that within the canopy most of Garstangetal.,1990;Fanetal.,1990).Subsequentprojects
the exchange of CO and H O happens on temporal scales in this region that kept the focus on this subject include
2 2
longer than 100s. At 80m, on the other hand, the turbu- theAnglo-BrazilianAmazonianClimateObservationStudy
lent fluxes are almost absent in such very stable conditions, (ABRACOS; Grace et al., 1995; Malhi et al., 1998; Kruijt
suggesting a boundary layer shallower than 80m. The rela- etal.,2000),theLarge-ScaleBiosphere-AtmosphereExper-
tionshipbetweenTKEandmeanwindsshowsthatthestable imentinAmazonia(LBA;Araújoetal.,2002;Saleskaetal.,
boundary layer switches from the very stable to the weakly 2003;Milleretal.,2004),ObservationsandModelingofthe
stableregimeduringintermittentburstsofturbulence.Ingen- Green Ocean Amazon (GOAmazon; Fuentes et al., 2016;
eral, fluxes estimated with long temporal windows that ac- Santos et al., 2016) and, most recently, the Amazon Tall
count for low-frequency effects are more dependent on the TowerObservatory(ATTO;Andreaeetal.,2015;Zahnetal.,
stabilityoveradeeperlayerabovetheforestthantheyareon 2016).
the stability between the top of the canopy and its interior, Ultimately, one of the most relevant questions that these
suggestingthatlow-frequencyprocessesarecontrolledover projectsaimedtoansweristheroleoftheAmazonrainfor-
adeeperlayerabovetheforest. est as either a net sink or source of CO2 to the atmosphere.
Results diverge greatly among the studies, from a net sink
of 5.9TCha−1yr−1 found by Grace et al. (1995) to a net
source of 1.3TCha−1yr−1 found by Saleska et al. (2003).
Although some of this variability may be accepted as gen-
uine,causedbysiteorinterannualdifferences,itisnowwell
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
3084 P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy
Figure1.Timeseriesofhorizontal(aandc)andvertical(e)windcomponents,temperatureperturbationfromthe20:00LTvalueat41m(b),
andCO2andO3(d)andwatervapor(f)concentrationsfortheturbulentnight.
Table1.Five-minuteturbulencestatisticsaveragedforeachnightanalyzedinSect.3.
14November2015 15November2015
intermittentnight turbulentnight
level σw (u(cid:48)w(cid:48)2+v(cid:48)w(cid:48)2)1/4 TKE σw (u(cid:48)w(cid:48)2+v(cid:48)w(cid:48)2)1/4 TKE
(m) (ms−1) (ms−1) (m2s−2) (ms−1) (ms−1) (m2s−2)
22 0.07 0.04 0.01 0.11 0.07 0.03
41 0.19 0.14 0.13 0.39 0.30 0.44
55 0.15 0.10 0.10 0.37 0.27 0.41
80 0.06 0.04 0.02 0.18 0.13 0.16
established that methodological problems affected the esti- small temporal scales (Vickers and Mahrt, 2006; Acevedo
matesthatfoundenhancedcarbonuptake.Mostoftheseis- et al., 2014). The exchange of properties such as CO from
2
suesregardthetreatmentofnocturnaldataasaconsequence theforesttotheatmospheremayoccurmostlythroughnon-
of the complex nature of the atmospheric flow during the turbulentmotion,suchasdrainageflows(StaeblerandFitz-
nightunder stable conditions.Inparticular,during verysta- jarrald,2004;Aubinetetal.,2003;Feigenwinteretal.,2004;
ble nights, turbulent mixing is reduced and constrained to Tóta et al., 2008) or by transport on temporal scales longer
Atmos.Chem.Phys.,18,3083–3099,2018 www.atmos-chem-phys.net/18/3083/2018/
P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy 3085
thanthosethatcharacterizetheturbulentflow(Santosetal., A comparison of scalar flux cospectra within and above
2016). aforestedcanopy,aimedspecificallyataddressingthecontri-
Themotionwithtemporalfluctuationslongerthanturbu- butionofnonturbulentflowtothetotalfluxesatthedifferent
lencebutsmallerthanthoseproducedbymesoscalesystems heights,hasnotbeenpresentedpreviously.Thepresentstudy
has been referred to as “submeso” by Mahrt (2009), and it aims at addressing this issue and to evaluate how these ex-
hasbecomeanimportantsubjectofmicrometeorologicalre- changeprocessesaffectthescalarprofilesthatareroutinely
search since then. Typically, these nonturbulent fluctuations measuredatATTO.
maybelargerinmagnitudethantheirturbulentcounterpart,
andtheymayintroducefluxesthatarelargeraswell.Onthe
other hand, these fluxes are not driven by local gradients, 2 Dataandmethods
so that they are also much more variable than the turbulent
fluxesandofeithersign,insuchawaythattheiroverallcon- 2.1 Experimentalsite
tribution frequently averages outover longer periods (Vick-
ersandMahrt,2003).Nevertheless,theircontributionmaybe ThedatasetwascollectedduringtheIntensiveOperatingPe-
important for closing the budgets over smaller time periods riod (ATTO-IOP1) in October/November 2015 at Reserva
(AcevedoandMahrt,2010;Kidstonetal.,2010). de Desenvolvimento Sustentável Uatumã (Uatumã Sustain-
Many studies on turbulence above and within forested ableDevelopmentReserve–USDR),intheAmazonregion.
canopies have presented an analysis of the spectral distri- Thesiteislocatedonaplateauat120ma.s.l.,approximately
bution of the turbulence velocity components and of their 150kmnortheastofManausand12kmnortheastoftheUa-
vertical variation with respect to the canopy top (Baldocchi tumãRiver.Theaverageheightofthehighesttreesnearthe
andMeyers,1988;Blankenetal.,1998;DupontandPatton, toweris37m.Furtherinformationregardingterrain,soil,and
2012). In general, these studies focused on the timescale of vegetationcanbefoundatAndreaeetal.(2015).
theturbulentexchangeandhowitdependsonfactorssuchas Micrometeorologicalobservationswerecarriedoutonan
distance from canopy top and atmospheric stability. Equiv- 80m walk-up tower with rectangular cross sections at five
alent analyses focusing on scalar flux cospectra have not different levels: 14, 22, 41, 55, and 80m above the ground.
been presented as often. Sakai et al. (2001) and Finnigan The first two levels are within the forest canopy, while
etal.(2003)usedcospectralsimilaritytoconcludethatlow- the three others are above it. Fast-response wind measure-
frequencycontributioncouldaccountformissingenergyand ments were performed at all levels (CSAT3, Campbell Sci-
CO fluxesintheirrespectivebudgets,butneitherstudyad- entific Inc., at 14, 41, and 55m; IRGASON, Campbell Sci-
2
dressed how the cospectra varied across the canopy. Other entificInc.,at22m;andWindmaster,GillInstrumentsLim-
studies (Campos et al., 2009; Fares et al., 2014) looked at ited, at 80m). Temperature has been measured by a pro-
scalar flux cospectra with the specific purpose of identify- file of thermohygrometers and by the sonic anemometers,
ing the proper temporal scale for turbulent flux determina- as well. The thermohygrometers have been intercompared,
tion. Santos et al. (2016) found that horizontal turbulent ki- whileeachsonictemperaturehasbeencomparedtotheclos-
netic energy (TKE) spectra are bimodal above an Amazo- estthermohygrometer.ScalarconcentrationsofCO andwa-
2
nianrainforestcanopy,withthepeakonshorttimescalesbe- tervaporweremeasuredat22m(IRGASON,CampbellSci-
ing related to turbulence and the peak on longer timescales entific Inc.), 41, and 80m (LI-7500A, LI-COR Inc.). The
being associated with nonturbulent, submeso fluctuations. diurnal cycle of the H O mixing ratios at 41m was erro-
2
Within the canopy, on the other hand, only the peak on neousforunidentifiedreasons.Theshort-term(upto20min)
longer timescales is preserved, indicating that nonturbulent variations were correct, but the longer trend agreed neither
fluctuationsabovethecanopypropagatedownwardmoreef- with the other open path instruments at 22 and 80m nor
ficiently than the turbulent ones. They also found that sen- with a nearby psychrometer (Frankenberger type, Theodor
sibleheatfluxcospectrawithinthecanopypeakedonlonger Friedrichs GmbH, Germany) or with the profile measure-
timescales,againsimilartothoseofthenonturbulentmaxima ments (see below). Therefore, the water vapor mixing ra-
of horizontal TKE above the canopy. This result indicates tios at that level haven been corrected by separating the
thattheexchangeofscalarswithinthecanopyatnightmay short-termfluctuationsfromthetrendbyapplyingarunning
occur on longer timescales than those traditionally used in mean with a window size of 5min and adding this high-
theeddycovarianceapproach.Underverystableconditions, frequency contribution to the running mean (window size
suchlongscalesmayalsocontributetothetotalexchangebe- 5min)ofthenearbypsychrometer.Scalarconcentrationsof
tween the canopy and the atmosphere. Santos et al. (2016), ozone were measured at 41m with chemiluminescence O
3
however, did not include the analysis of CO or latent heat sondes (Enviscope, Germany). In front of the fast O in-
2 3
fluxes and reactive trace gases like O , so that the question strument there was a 5m long 3/4(cid:48)(cid:48) (7.52mm inner diam-
3
ofwhetherthesequantitiesareaffectedbysimilarprocesses eter) Teflon tubing with a Teflon inlet filter (47mm diame-
remainsopen. ter, 5µm pore size). The flows varied due to filter clogging.
Afterafilterchangetheflowswere21and23.5Lmin−1,re-
www.atmos-chem-phys.net/18/3083/2018/ Atmos.Chem.Phys.,18,3083–3099,2018
3086 P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy
Figure2.SameasFig.1butfortheintermittentnight.Shadedareasindicateintermittentturbulencebursts.
Figure3.ConcentrationsofCO2(a)andO3(b)asafunctionoftimeandheightfortheturbulentnight.
spectively,whereasbeforethefilterchangetheywere16and theresidencetimewastherefore0.8s.Allthedatawerecol-
14Lmin−1, respectively. The resulting lag times were 0.6– lected at a rate of 10Hz. As the signal of the fast O son-
3
0.95s,andtheReynoldsnumbersinthetubingwere2400to des has considerable drift, it was calibrated to a slow O
3
4000 at 35◦C. On the days considered for the case studies analyzer (TEI 49i, Thermo Scientific) as described by Zhu
(14 and 15November),the flow was about 16Lmin−1, and et al. (2015). The CO profiles were measured sequentially
2
Atmos.Chem.Phys.,18,3083–3099,2018 www.atmos-chem-phys.net/18/3083/2018/
P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy 3087
Table2.Five-minuteTKEandfluxesaveragedforshadedandnon-shadedareasintheintermittentnight(seeFig.2).
14November2015–intermittentnight
Level FC FH Fq TKE FO
(m) µmolm2s−1 Wm−2 Wm−2 (m2s−2) nmolm2s−1
Shaded 22 1.6 −2.9 3.7 0.01
41 2.6 −15.9 9.6 0.16 −1.5
55 – −12.5 – 0.13
80 0.0 −0.4 0.9 0.02
Non-shaded 22 1.6 −0.4 3.7 0.01
41 1.0 −3.2 2.9 0.06 −0.4
55 – −2.4 – 0.03
80 0.0 −0.1 0.4 0.02
by CO /H O analyzers (LI-7000, LI-COR Inc.) connected considered for this analysis. For this reason, nocturnal peri-
2 2
to heated inlets at eight heights (0.05, 0.5, 4.0, 12.0, 24.0, odswererestrictedtothetimefrom20:00to05:00LT.Since
38.3,53.0,and79.3m).Duringthecasestudynights(14and thedifferentlevelsofflowstructuresareanalyzedsimultane-
15 November), only the LI-7000 behind the Nafion® dryer ously,onlythedatawhenalllevelswereavailablewereused.
wasrunning,andthereforethewatervaporvaluescouldnot The 14m level frequently presented gaps and was not con-
beused.TheO profileswerealsomeasuredusingthesame sideredforthisstudy.Therefore,thelevelsincludedareone
3
inlet system with an ozone analyzer (TEI 49i, Thermo Sci- insidethecanopy(22m),onejustabovecanopytop(41m),
entific).Ambientairwascontinuouslydrawnfromtheinlets andtwowellabovethecanopy(55and80m).
throughnontransparentPTFEtubing(3/8(cid:48)(cid:48))toavalveblock, All the time series have been subject to quality control,
whichswitchedbetweenthedifferentinletlevels,sothatone whichcausedtheremovalofthoseserieswhichshowedmul-
intakeheightwaspurgedbythesamplepump(PTFEcoated), tiple spikes or multiresolution spectra that displayed noise
whilealltheotherswerepurgedbythebypasspump.Atime ontheshortesttimescales.Foranycasewhereagivenseries
intervalof1minwasnecessaryforgettingaconstantandre- was discarded for a given variable, it has not been used for
liable signal for each concentration level: a complete cycle any of the variables. With these restrictions, 15 nights were
took8×2=16min,providingtwomeasurementsperlevel. kept for the final analysis. Ozone measurements started on
Three 16min measurement cycles plus one shorter 12min 11 November 2015, so that only nine nights of ozone flux
cycle were performed every hour. During that last cycle, datawereavailable.
asmallcompromisewasmadetofitfourcyclesintothehour, Thedatawereanalyzedusingtwodifferenttimewindows:
andvalveswitchesoccurredevery90s,therebyallowingfor 5 and 109min. The multiresolution decomposition (Howell
onlyoneconcentrationmeasurementateachlevel.Theam- andMahrt,1997;VickersandMahrt,2003;Voronovichand
bient air inlets mounted on the tower were protected from Kiely, 2007) was applied to 109min, which corresponds to
rainenteringtheinletlinebypolyethylenefunnelsandfrom groups of 216 data points. In contrast to the Fourier trans-
insects by polyethylene nets. A PTFE filter (5µm) was in- form, which determines periodicity, this technique mainly
stalled right after the inlet. The tubing was insulated with extractsthewidthofthedominantturbulenteventsbylocally
Styrofoamandheated.Theinternaltemperatureandpressure decomposingthevariances.Forthisreason,themultiresolu-
correctionsoftheLI-7000wereused,buttofurtherminimize tionspectrum(S)andcospectrum(C)havethepropertythat
pressureeffects,theairdrawnfromtheinletsforanalysiswas theintegrationuptoagiventimescaletisequaltothevari-
sampledattheexitoftheTeflonpump,sothatthemeasure- ance and covariance, respectively, for a t-long time series.
mentsweremadeclosetoambientpressureforallmeasure- Consequently,themultiresolutionvalueforagiventimescale
ment levels. The entire setup was comparable to the profile capturesthephysicalprocesses(andtheflux)whoseduration
systememployedbyMayeretal.(2011). issmallerthanthattimescale.
The multiresolution decomposition was applied sequen-
2.2 Dataanalysis tiallytothetimeseries,startingat20:00LT,withanoverlap
of30minbetweenthesubsequentseries,totaling14decom-
In the present study, 20 days of nocturnal data were ana- positionsforeachnight.Atotalof200serieswasusedinthe
lyzed, from 1 to 20 November 2015. To avoid sampling in- study,consideringallnights.AcevedoandMahrt(2010)used
tense events associated with the transitional characteristics the multiresolution decomposition to analyze vertical pro-
betweendaytimeandnocturnalboundarylayers,theevening files of the nonturbulent component of sensible heat fluxes.
periodbetweenthesunsetand20:00localtime(LT)wasnot Theyfoundthatsystematicandorganizedprofiles,whosein-
www.atmos-chem-phys.net/18/3083/2018/ Atmos.Chem.Phys.,18,3083–3099,2018
3088 P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy
Figure4.SameasFig.3butfortheintermittentnight.
Figure5.Fluxesofsensibleheat(aandb),CO2(candd),O3(eandf),andlatentheat(gandh)fortheturbulentnight(a,c,e,g)andfor
theintermittentnight(b,d,f,h).ShadedareasindicateintermittentturbulenceburstsasshowninFig.2.
clusion contributes to the closure of the nocturnal tempera- cedures, such as globally directionally dependent methods
turebudgetnearthesurface,areonlyfoundwhenthedouble (Lee,1998;Mahrtetal.,2000;PawUetal.,2000)because
windrotation(TannerandThurtell,1969)isappliedtoeach “the measured vertical motion on times scales greater than
timeseriesanalyzed,separately.Theyclaimthatthisismore 5h may be sufficiently weak and unreliable that the elimi-
suitableforsuchanalysisthanothercoordinaterotationpro- nation of larger-scale variations in vertical motion through
Atmos.Chem.Phys.,18,3083–3099,2018 www.atmos-chem-phys.net/18/3083/2018/
P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy 3089
Table3.TKEandfluxesaveragedforeachnightanalyzedinSect.3usingatimewindowof5and109min.
14November2015 15November2015
intermittentnight turbulentnight
Level FC FH Fq TKE FO FC FH Fq TKE FO
(m) µmolm2s−1 Wm−2 Wm−2 (m2s−2) nmolm2s−1 µmolm2s−1 Wm−2 Wm−2 (m2s−2) nmolm2s−1
109min 22 4.0 −2.0 8.8 0.02 1.1 −2.1 2.8 0.03
41 1.2 −7.9 22.9 0.29 −0.9 3.5 −37.4 29.7 0.49 −2.3
55 – −2.4 – 0.48 – −32.4 – 0.49
80 −0.1 1.0 −2.4 0.47 3.9 −13.4 20.4 0.29
5min 22 1.6 −2.1 3.7 0.01 1.0 −1.3 2.6 0.03
41 2.1 −11.9 7.5 0.13 −1.2 3.8 −35.6 29.4 0.44 −1.3
55 – −9.3 – 0.10 – −30.2 – 0.41
80 0.0 −0.3 0.7 0.02 4.3 −14.1 20.8 0.16
canopy at another site in the Amazon forest occurs on tem-
poral scales smaller than 200s. The use of a 109min long
time window is necessary to determine the contribution of
turbulent and nonturbulent motions to the fluxes. However,
inordertoattempttoreduceanycontributionfromnonturbu-
lenttransport,statisticalmoments,fluxes,andothervariables
werealsocalculatedusinga5mintimewindow,asusedby
Dupont and Patton (2012). Quantities such as sensible heat
(F =w(cid:48)θ(cid:48)), latent heat (F =w(cid:48)q(cid:48)), CO (F =w(cid:48)C(cid:48)),
H q 2 CO2
ozone(F =w(cid:48)O(cid:48))fluxes,turbulentkineticenergy(TKE),
O3
the average horizontal wind speed (V), Richardson num-
ber (Ri), and σw were determined for both 5 and√109min.
The turbulent velocity scale, defined as V = TKE=
TKE
[0.5(σ +σ +σ )]1/2, was calculated for 5min time win-
u v w
dows only. This study comprises a total of 1577 5min win-
dows.
Figure6.AverageTKEspectrafortheturbulentnight(blacksolid The bulk Richardson number was used to quantify atmo-
lines)andtheintermittentnight(reddashedlines)foralllevels,as spheric stability. The choice of the bulk instead of the flux
indicated in each panel. In all panels, averages are calculated for Richardsonnumberfortheanalysishastworeasons:toavoid
eachtimescale. self-correlation (Hicks, 1978; Klipp and Mahrt, 2004; Baas
et al., 2006) and to quantify better the stability in very sta-
bleconditionswhenfluxesareexpectedtoapproach0.Sim-
coordinaterotationimprovesthecalculation”.Fortheserea- ilarly to usage in Bosveld et al. (1999), Mammarella et al.
sons, the double rotation was applied to each 109min time (2007),andOliveiraetal.(2013),a“within-canopyRichard-
seriesseparately. sonnumber”(Ri )andan“above-canopyRichardsonnum-
can
Variances and fluxes with a 109min long time average ber”(Ri )(Santosetal.,2016)weredefinedas
top
were obtained from the integration of the respective mul-
tiresolution spectra and cospectra for each series separately g θ −θ
Ri = (cid:49)z 41m 22m (1)
and then averaged, if appropriate. Therefore, sensible and can (cid:50) (V −V )2
41m 22m
latent heat, CO , and ozone fluxes are given by F =
2 H
PτCw(cid:48)θ(cid:48), Fq =PτCw(cid:48)q(cid:48), FCO2 =PτCw(cid:48)C(cid:48), and FO3 = and
P C ; turbulent kinetic energy is TKE=0.5P (S +
τ w(cid:48)O(cid:48) τ u
Sv+Sw) and the SD of the vertical wind component is Ri = g(cid:49)z θ80m−θ41m , (2)
σw=(PτSw)1/2. Other variables, such as the Richardson top (cid:50) (V80m−V41m)2
number (Ri) and average horizontal wind speed (V), were
calculatedusingthesametimeintervalusedinthemultires- where g is the gravitational acceleration, (cid:50) is the average
olutiondecomposition. potentialtemperatureinthelayer,(cid:49)zistheheightdifference
Atnight,itisexpectedthatthetemporalscalesofturbulent betweenthetwolevels,andθ andV arethemeanpotential
transport are smaller. Campos et al. (2009) showed that the temperatureandaveragehorizontalwindspeedateachlevel,
contribution of turbulence to the nocturnal fluxes above the respectively.
www.atmos-chem-phys.net/18/3083/2018/ Atmos.Chem.Phys.,18,3083–3099,2018
3090 P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy
Figure7.Averagecospectraofsensibleheat(a,d,f,i),CO2(b,g,j),O3(e),andlatentheatfluxes(c,handk)fortheturbulentnight(black
solidlines)andtheintermittentnight(reddashedlines)foralllevels,asindicatedineachpanel.Inallpanels,averagesarecalculatedfor
eachtimescale.
3 Comparisonofturbulencecharacteristicsinafully the pollutant dispersion community similar phenomena are
turbulentwithanintermittentlyturbulentnight oftenreferredtoasmeandering(Oettletal.,2005).Themost
relevantdifferencebetweenthetwonightsregardsthemag-
nitude of the turbulent mixing (Table 1). All relevant turbu-
The nocturnal flow at the site is characterized by the super-
lencestatisticsaresignificantlylargeron15Novemberthan
positionofturbulentandnonturbulentfluctuations.Inafully
on 14 November. The relative difference of the turbulence
turbulentnight,suchas15November2015(Fig.1),thereis
statisticsbetweennightsincreasessteadilyinthevertical.As
a clear dominant wind direction at all levels. In this case, it
an example, TKE at 41m is 3.4 times larger in the turbu-
isveryrarethatthehorizontalwindcomponentsswitchsign
lentnightthanintheintermittentcase,whileat80m,TKEis
abovethecanopy.Incontrast,duringtheintermittentnightof
8.2timeslargerintheturbulentnight.Similarincreasesoc-
14November2015(Fig.2),thereisnodominantwinddirec-
tion at any level above the canopy, as both horizontal com- curforthecorrespondingratiosofσwand(u(cid:48)w(cid:48)2+v(cid:48)w(cid:48)2)1/4
ponentsswitchsignmanytimesthroughoutthenight.Low- betweenthetwonights.
frequencyfluctuationsaresuperposedontheturbulentfluctu- Anotherinterestingcharacteristicthatindicatesacontrast
ations,causingthemeanwinddirectiontochangequadrants betweenthetwonightsshowninFig.1andFig.2regardsthe
frequentlythroughoutthenight.Suchfluctuationshavebeen degreeofverticalcouplingacrossthelevels,aphenomenon
recently attributed to submeso flow (Mahrt, 2009), while in thathasbeenobservedbyvanGorseletal.(2011),Oliveira
Atmos.Chem.Phys.,18,3083–3099,2018 www.atmos-chem-phys.net/18/3083/2018/
P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy 3091
Figure8.AveragespectraofTKE(a)andcospectraofCO2andO3(b),sensibleheat(c),andlatentheat(d)fluxesfortheentiredataset.In
allpanels,averagesarecalculatedforeachtimescale.
etal.(2013),andJocheretal.(2017).Intheturbulentnight, eventsofcouplingoccurredduringburstsofintermittenttur-
temperatures were always similar between the levels of 41 bulence (Fig.2, shadedareas). It hasbeen assumedthat the
and55m,whileat80m,itwasslightlywarmer,butwiththe turbulentburstsoccurredwheneverσ >0.15ms−1.During
w
samecoolingtendencythroughouttheperiod.CO wascor- these events of coupling, the gradients of temperature and
2
respondingly similar between 22 and 41m, with the same CO concentration became sporadically smaller across the
2
tendencies and slightly lower values at 80m. Although the vertical, except for the 80m level, indicating that the cou-
meantrendissimilarat22and41m,substantial,shorttime plinginducedbytheeventsextendedoveralayershallower
deviationstowardshigherCO valueswereobservedat22m than 80m. In general, the temporal evolution of all scalars
2
(Fig. 1d). This is in line with the higher variability in CO shows a monotonic increase (in CO ) or decrease (in tem-
2 2
valuesinthelowercanopy,ascanbeseenfromtheprofiles peratureandO )throughouttheturbulentnightatalllevels
3
(Fig. 3a). This higher variability and stronger gradient (in (Fig.1).Intheintermittentnight,ontheotherhand,largein-
bothCO andO )inthelowercanopypointtoadecoupling creasesanddecreasesinallscalarsoccurinsmallperiodsof
2 3
of the sub-canopy even in the turbulent night. As the 22m timeatalllevels,exceptat80m.AsCO hasaclearsource
2
level is within the maximum of the leaf area index (LAI) atthegroundandO hasaclearsinkattheground,onecan
3
(∼24m), which separates the upper canopy from the lower identify from the profiles whether air is coming from aloft
canopy,itwillbeinfluencedbybothregimes.Thegradients or from below (Fig. 4). Air from above is rich in O and
3
between 24 and 38m are always positive for O and nega- lower in CO , whereas air from below is rich in CO but
3 2 2
tive for CO . This can be related to the reactivity of O as depletedinO .Fromthisperspective,inthefirsteventairis
2 3 3
itreactswithcompoundsemittedfromthesoil(mainlyNO) mixeddownfromaloft,whileinthesecondeventairismixed
andplants(alkenes)andisnotonlytakenupbystomatabutis bothupwardanddownwardfromthecanopytop.Inthethird
alsodepositedtoleafsurfacesinconsiderableamounts,espe- event, air is first mixed down and finally there is a burst of
ciallyunderhumidconditions(FuentesandGillespie,1992; air going upwards from the canopy. At 80m, temperature
Rummeletal.,2007).Atnight,CO isemittedbysoilsand (Fig.2b)andCO (Fig.2d)showmuchsmallerfluctuations
2 2
plantsduetorespiration,causinganegativegradient. thanattheotherlevels.Thisisfurtherevidencethatthesta-
Allquantitiesshowedmuchlargervariationacrossthelev- ble boundary layer (SBL) thickness is shallower during the
els in the intermittent night (Fig. 2). Furthermore, sporadic intermittent night, such that the canopy exchange fluxes do
www.atmos-chem-phys.net/18/3083/2018/ Atmos.Chem.Phys.,18,3083–3099,2018
3092 P.E.S.Oliveiraetal.:NighttimewindandscalarvariabilitywithinandaboveanAmazoniancanopy
Figure9.Thedependenceofsensibleheat(a,d,f,i),CO2(b,g,j),O3(e),andlatentheat(c,handk)fluxesoncanopyRichardsonnumber
(Rican) for all levels, as indicated in each panel. Fluxes have been determined with 5min (red lines, triangles) and 109min (black lines,
circles)timewindows.
notaffectthestateoftheatmosphereat80m.Thisfactcon- the longest timescales are the most energetic, but the 10s
trastsstronglywiththesteadytrendsinbothscalarsat80m turbulentmaximumandacospectralgap(near100s)arestill
during the turbulent night (Fig. 1b and Fig. 1d), which in- evident.Incontrast,intheintermittentnightof14November,
dicates that in this case, this level is fully coupled through at all levels most energy prevails on the longest timescales
turbulence to the canopy top. While in the turbulent night- providedbythedecompositionmethod.Thisenergyisasso-
time, scalar fluxes did not vary substantially throughout the ciatedwiththelow-frequencyfluctuationsresponsibleforthe
period(Fig.5a,c,e,andg),themostintenseturbulentfluxes variabilityinthewinddirectionvisibleinFig.2aandFig.2c.
of sensible heat (Fig. 5b), CO (Fig. 5d), O (Fig. 5f), and These spectra confirm that when fully turbulent conditions
2 3
latent heat (Fig. 5h) during the intermittent night occurred prevail,theenergyofthenonturbulent,low-frequencymodes
duringthesecouplingperiods(Table2). of the flow is reduced considerably. The cospectra of the
Previous studies have reported that nonturbulent modes fluxes of sensible heat (Fig. 7a,d,f, and i), CO (Fig. 7b,g,
2
of the flow only become relevant when turbulence is weak and j), O (Fig. 7e), and latent heat (Fig. 7c,h, and k) con-
3
(Acevedoetal.,2014),alikelyconsequenceofthediffusive firmtheenhancedturbulentexchangeofallquantitiesinthe
nature of turbulence destroying the nonturbulent temporal turbulent night compared to the intermittent one. They also
andspatialvariabilityintheatmosphericvariables.Thisrela- show that, consistently to what occurs with TKE, the non-
tionshipbetweentheturbulentandnonturbulentmodesofthe turbulentexchangeofthesescalarsisenhancedintheinter-
flow is illustrated by the TKE spectrum during both nights mittent case. In particular, a significant low-frequency flux
(Fig. 6). In the turbulent case, most of the energy is associ- of CO occurs at 22m in the intermittent night, such that
2
atedwithturbulence,sothatthemostenergetictimescaleis the total flux at this level is larger during the intermittent
near10satalllevels,exceptfor80m(Fig.6a).Atthislevel, night(4.0µmolm2s−1,Table3)thanduringtheturbulentone
Atmos.Chem.Phys.,18,3083–3099,2018 www.atmos-chem-phys.net/18/3083/2018/