Plants remember past weather: a study for atmospheric pollen concentrations of Ambrosia ...

13 Pages·2015·0.84 MB·English

TheorApplClimatol(2015)122:181–193
DOI10.1007/s00704-014-1280-2
ORIGINALPAPER
Plants remember past weather: a study for atmospheric pollen
Ambrosia Populus
concentrations of , Poaceae and
IstvánMatyasovszky&LászlóMakra&ZoltánCsépe&
ZoltánSümeghy&ÁronJózsefDeák&
ElemérPál-Molnár&GáborTusnády
Received:12February2014/Accepted:15September2014/Publishedonline:9October2014
#Springer-VerlagWien2014
Abstract After extreme dry (wet) summers or years, pollen aswellasdailyvaluesoffourclimatevariables(temperature,
productionofdifferenttaxamaydecrease(increase)substan- relative humidity, global solar flux and precipitation).
tially.Accordingly,studyingeffectsofcurrentandpastmete- Multiplecorrelationsofdailypollencountswithsimultaneous
orological conditions on current pollen concentrations for values ofdailymeteorologicalvariables donot showannual
differenttaxahaveofmajorimportance.Thepurposeofthis courseforAmbrosia,butdoshowdefinitetrendsforPopulus
study is separating the weight of current and past weather andPoaceae.Resultsreceivedusingthetwomethodsrevealed
conditions influencing current pollen productions of three characteristicsimilarities.Forallthethreetaxa,thecontinental
taxa.Twoprocedures,namelymultiplecorrelationsandfactor rainfall peak and additional local showers in the growing
analysis with special transformation are used. The 11-year seasoncanstrengthentheweightofthecurrentmeteorological
(1997–2007) data sets include daily pollen counts of elements. However, due to the precipitation, big amount of
Ambrosia(ragweed),Poaceae(grasses)andPopulus(poplar), watercanbestoredinthesoilcontributingtotheeffectofthe
past climate elements during dry periods. Higher climate
sensitivity(especiallywatersensitivity)oftheherbaceoustaxa
I.Matyasovszky
(Ambrosia and Poaceae) can be definitely established com-
DepartmentofMeteorology,EötvösLorándUniversity,
paredtothearborealPopulus.Separationoftheweightofthe
1117BudapestPázmányPéterst.1/A,Hungary
e-mail:[email protected] currentandpastweatherconditionsfordifferenttaxainvolves
: : practical importance both for health care and agricultural
L.Makra(*) Z.Csépe Z.Sümeghy
production.
DepartmentofClimatologyandLandscapeEcology,Universityof
Szeged,6701SzegedP.O.B.653,Hungary
e-mail:[email protected]
Z.Csépe
e-mail:[email protected] 1Introduction
Z.Sümeghy
e-mail:[email protected] Allergic diseases have become one of the most important
health issues worldwide (Bocking et al. 2012; de Marco
Á.J.Deák et al. 2012; D’Amato et al. 2013). Sensitisation and allergic
DepartmentofPhysicalGeographyandGeoinformatics,University
diseases increase with socioeconomic status and living in
ofSzeged,6701SzegedP.O.B.653,Hungary
e-mail:[email protected] large cities (Langen et al. 2013). Allergy and asthma, espe-
cially those associated with ragweed pollen, have been pre-
E.Pál-Molnár vailinginHungary(Kazinczi etal. 2008;Páldy etal. 2010).
DepartmentofMineralogy,GeochemistryandPetrology,University
Recently, 20 % of the Hungarian population suffers from
ofSzeged,6701SzegedP.O.B.653,Hungary
e-mail:[email protected] pollen allergy and 60-90 % of them are sensitised with rag-
weed pollen (Harsányi 2009). In 1998–1999, according
G.Tusnády to skin prick tests, 83.7 % of the patients were sensitive
MathematicalInstituteoftheHungarianAcademyofSciences,
to ragweed pollen in Szeged, Hungary (Kadocsa and
1364BudapestP.O.B.127,Hungary
e-mail:[email protected] Juhász 2002).
182 I.Matyasovszkyetal.
The release of allergenic pollens depends both on phyto- Although several studies have been published to explore
physiological status of the given taxon and meteorological therelationshipbetweenmeteorologicalconditionsandpollen
conditions. There is a vast amount of literature dealing with loads, as well as past weather conditions and certain pheno-
therelationshipbetweendailyallergenicpollenconcentrations logicalphasesofdifferenttaxa,neitherofthemwasaimedto
and simultaneous weather conditions (e.g. Bartková- distinguishbetweentheeffectofcurrentandpastweatheron
Ščevková 2003; Kasprzyk 2008; Veriankaitè et al. 2011; currentpollenconcentrations.Thepurposeofthispaperisto
Prtenjak et al. 2012). The current pollen concentration is, separatetheweightofthecurrentandpastclimateconditions
however,affectednotonlybythecurrentweatherconditions in determining the pollen concentrations of Ambrosia,
butalsobytheeffectsoftheantecedentmeteorologicalcon- Poaceae and Populus for the Szeged region in the Southern
ditions duringa period. Taxa withshort pollenseasons have Hungary applying two procedures, namely multiple correla-
longerperiodofpastweatherconditionsbetweenthelastand tionsandfactoranalysiswithspecialtransformation.
theforthcomingpollenseasonsandviceversa.
Theinfluenceofpastweatherconditionsoncertainpheno-
logical phasesofdifferenttaxahas beeninvestigatedinsev- 2Materialsandmethods
eral studies. Emberlin and Norris-Hill (1991) found that an-
nualdifferencesinthecumulativeUrticaceaepollenconcen- 2.1Locationanddata
trationwereprimarilyduetoweatherconditionsintheperiod
of pollen formation and only secondarily due to weather Szeged (46.25 N; 20.10 E), the largest settlement in the
conditionsinthepollenreleaseseason.Furthermore,relative southeastern Hungary is located at the confluence of the
humidity, temperature, wind speed and rainfall were most RiversTiszaandMaros(Fig.1).Theareaischaracterisedby
important in daily variations, but their relative importance an extensive flat landscape of the Great Hungarian Plain,
varied between years. Spieksma et al. (1995) found that the namely Pannonian Plain, with an elevation of 79 m above
air temperature during the preceding 40 days has a decisive sea level. The city is the centre of the Szeged region with
influenceonthestartdateoftheBetulapollenseason.Giner 203,000 inhabitants. The climate of Szeged belongs to
etal. (1999) related daily Artemisia pollenconcentrationsto Köppen’s Ca type (warm temperate climate) with relatively
therainfallandglobalsolarfluxintheprecedingweeks.They mildandshortwintersandhotsummers(Köppen1931).
foundthatoncepollinationhadbegun,meteorologicalfactors Thepollencontentoftheairwasmeasuredusinga7-day
(excluding wind direction) did not seem to influence pollen recording Hirst-type volumetric spore trap (Hirst 1952)
concentrationssignificantly.Linderholm(2006)foundthatthe (Fig.1).The air sampler islocated ontopofthe building of
timingofspringeventsinmid-tohigh-latitudeplants,suchas theFacultyofArtsattheUniversityofSzeged,approximately
budding, leafing and flowering, is mainly regulated by tem- 20mabovethegroundsurface(Makraetal.2010).
peratures after the dormancy is released. For breaking dor- Dailyvaluesofthecurrentandprecedingweathervariables
mancy, chilling temperatures (temperatures that break dor- [meantemperature,(T,°C)],relativehumidity(RH,%),global
mancy) accelerate bud growth from the state of quiescence solar radiation (GSR, Wm−2) and precipitation amount (P,
(i.e.whendormancyisbroken)tothestateofburst(Hänninen mm) influencing daily pollen concentrations were used in
etal.1993;Chuine2000).Namely,themorechillingtemper- the study. They were collected in a meteorological station
atures are received, the less forcing temperatures (tempera- located in the inner city area of Szeged (Fig. 1). These ele-
turesthatforcegrowthduringspringwhendormancyhasbeen ments were used since they indicate the highest impact on
released)aresubsequentlyneededtoreachbudburst(Chuine pollen production (Galán et al. 2000; Bartková-Ščevková
2000).Therefore,higherwinterandearlyspringtemperatures 2003; Štefanič et al. 2005; Kasprzyk 2008; Makra and
can produce a later onset of bud burst and pollen release. Matyasovszky 2011). The data set consists of daily pollen
Emberlin et al. (1997) found a trend to earlier Betula pollen counts(averagedailypollencountpercubicmetreofair)of
seasonsthatwasrelatedtoanincreaseincumulativetemper- three taxa taken over their longest common measurement
atures over 5.5 °C in January, February and March. period, namely the 11-year period 1997–2007. With their
Furthermore,theydetectedasignificantpositiverelationship Latin (English) names, they are: Ambrosia (ragweed),
between the start dates of the Corylus pollen season and Poaceae (grasses) and Populus (poplar). These three taxa
temperatures in October suggesting that lower temperatures releasethehighestpollenlevels,namelyAmbrosia(32.6%),
inOctoberresultinanearlieronset,andviceversa(Emberlin Poaceae (10.4 %) and Populus (9.2 %) together account for
et al. 2007). They found a significant negative association 52.2%ofthetotalpollenproductioninSzeged(Table1).
betweenonsetandDecember temperatures indicating that Selection of these taxa was due to their high or medium
higher December temperatures will produce an earlier allergenicity [in a four-score scale found in the Hungarian
onset of Corylus flowering and vice versa (Emberlin pollen index (www.pollenindex.hu), allergenicity of both
et al. 2007). AmbrosiaandPoaceaeisthehighestindicatedbyscorefour,
Plantsrememberpastweather 183
Fig.1 LocationofEurope
includingHungary(upperpanel)
andtheurbanwebofSzegedwith
thepositionsofthedatasources
(lowerpanel).Number1
indicatesthemeteorological
stationand2theaerobiological
station.Thedistancebetweenthe
aerobiologicalandthe
meteorologicalstationis2km
whilethatofPopulusismediumindicatedbyscoretwo]and Anotheraspectoftheselectionwastoanalysetwoherbaceous
totheirmoreorlesspermanentlyhighpollenconcentrations. taxa(AmbrosiaandPoaceae)andanarborealdeciduoustaxon
(Populus).
Table1 Planthabitsandphenologicalpollenseasoncharacteristics
RagweedisahighlyprevalenttaxonnotonlyintheSzeged
Taxa Planthabit aPollenlevel Pollenseason area but also over the Pannonian Plain in the Carpathian
Basin,aswell.Theyoccurinstubblefields,especiallyinsand
Start End Length,day
landscapesandinabandonedplacesaroundsettlements.The
increaseofmeantemperatureforthewarm-tolerantAmbrosia,
Ambrosia H 32.6 Jul15 Oct15 93
especiallyinsummertime(August),leadstoalimitforpollen
Poaceae H 10.4 Apr16 Oct11 180
production(SzigetváriandBenkő2004;Béresetal.2005).In
Populus AD 9.2 Feb27 Apr20 53
this period, in order to preserve water, it decreases pollen
ADarborealdeciduous,Hherbaceous production. Higher potential distribution of Ambrosia is an-
aInpercentageofthetotalpollendispersion ticipatedduetoitshighclimatetolerance.Namely,thisgenus
184 I.Matyasovszkyetal.
canadaptwelltodryandhotconditions.Ifmorefallowareas RH,GSRandP.ThesecondgroupconsistsofcumulativeT,
and abandoned human habitats appear in the landscape, its RH, GSR and P, which are defined asfollows.LetD bethe
furtherincreaseisexpected,especiallyonsandysoils(Makra durationofthepollenseasonofagiventaxon,andletdbethe
etal.2011;Deáketal.2013). durationoftheperiodfromthefirstpollen-freedayfollowing
Bothelderfallowsandnaturalgrasslandsarecharacterisedby thepreviouspollenseasontothelastpollen-freedaypreced-
ahugecoverageofgrasses,sofallowregenerationintheSzeged ingtheactualpollenseason.Ontheithdayofanactualpollen
regionledtotheextensionofgrass-coveredareas.Poaceaecan season,valuesofthemeteorologicalvariablesareaccumulat-
produce high biomass in years with higher than usual rainfall, edfromthe(i−d+j)thdaytothe(i−1)thdaycorrespondingto
which is in accordance with their higher pollen production. At accumulationlengthsd−jfori=1,…,D.Thelinearregression
thesametime,increasedmeanandmaximumtemperaturescan is carried out for every day of the pollen season (i=1,…,D)
causewatershortageinthedriestsummerperiodalsoforgrasses, with every accumulation length d−j (j=0,…,d−1). For in-
meaningaseriouslimitingfactor.Sotheypreservewaterinstead stance,PoaceaeexhibitsapollenseasoninSzegedfromApril
of producing pollen. Poaceae show high climate sensitivity. 16toOctober12(Table1)andthusD=180andd=186days.
However, the species pool of this family is the widest in Adjustedmultiplecorrelations (e.g. Draperand Smith 1981)
Hungary among the three taxa analysed in the study, so there for both groups of explaining variables are calculated for
will be other species to substitute the actual grasses, and even everydayofthepollenseasonandeveryaccumulationlength
species from both the Mediterranean and continental areas can in the past. The procedure outlined here is applied to the
reachtheCarpathianBasininthefuture(Deák2010;Makraetal. above-mentionedthreetaxa.
2011;Deáketal.2013).
Populus indicate a wide climate tolerance since both wet 2.3Factoranalysiswithspecialtransformation
anddrytolerantspeciesarerepresentedinthelandscapefrom
the floodplains to the bare sand. Populus (both wild and Factoranalysisidentifieslinearrelationshipsamongexamined
cultivated types) were planted widely especially in the sand variablesandthushelpstoreducethedimensionalityofalarge
lands,westfromSzegedcity,aswellasinthefloodplainstoo. datasetofp-correlatedvariables,expressingthemintermsof
Plantationofthesespecieshasnotyetstoppedduringthelast m (m<p) new uncorrelated variables, the so-called compo-
10years.Besidesthelocusttree(Robiniapseudoacacia),they nents or factors. Calculation was based on PCA combined
are the most favoured trees for plantation of forests. The with varimax rotation keeping the factors uncorrelated
stocksplantedinthelastdecadeshavegrownup;theyarein (Sindosi et al. 2003). The number of components (factors)
maturestate,sotheycanpollinateonhighlevel.Thewarmer produced is equivalent to the number of original input vari-
andmoderatelyhumidweatherinthespringalsofavourstheir ablesandaccountfor100%ofthetotalvarianceofalloriginal
pollination(Deák2010;Makraetal.2011;Deáketal.2013). variables.Sinceonlyafewcomponentsmayaccountforthe
Thepollenseasonisdefinedbyitsstartandenddates.For majorityofthetotalvariance,itmaybeunnecessarytoretain
the start (end) of the season, we used the first (last) date on allcomponents.Severalmethodsareavailablefordetermining
which1pollengrainm−3ofairisrecorded,andatleastfive thenumberofcomponentstoberetained(Jolliffe1990,1993;
consecutive (preceding) days also show 1 or more pollen McGregor and Bamzelis 1995). One of the most known
grains m−3 (Galán et al. 2001). For a given pollen type, the component selection techniques selects components with ei-
longest pollen season during the 11-year period was consid- genvalues>1.Thisisbasedontheideathatallnewvariables
eredforeachyear. shouldhavegreaterexplanatorypowerthantheoriginalvar-
Dailyvaluesofmeantemperature,relativehumidity,global iables, which have an eigenvalue of 1. Some papers have
solar radiation and precipitation amount were labelled as established that selecting components with eigenvalues <1
current meteorological conditions for the pollen season of a can result in an increase of the explanatory power and thus
giventaxon,whiletheywereconsideredaspastmeteorolog- suggestedretainingthenumberofcomponentswiththelarg-
icalconditionsfortheperiodstartingfromthefirstpollen-free est explained cumulative variance that account for at least
day following the previous pollen season to the last pollen- 80%ofthetotalvarianceofthe originalvariables thathave
freedayprecedingtheactualpollenseason. tobeexplainedbythefactors(Jolliffe1990).Notethatthese
methods are consideredsubjective (McGregor and Bamzelis
2.2Regression 1995). We applied the latter procedure, as perhaps the most
commonmethod,forcalculatingthenumberofcomponentsto
Linear regressions are used in our present study as follows. beretained(Sindosietal.2003;Liu2009).Factoranalysiswas
The explaining variables are divided into two parts, and the applied to our initial data sets consisting of nine correlated
daily pollen concentrations are regressed separately against variables(eightexplanatoryvariablesincludingfourmeteorolog-
thesetwogroupsofmeteorologicalvariablesforeverydayof ical variables characterising the weather of actual days and the
thepollenseason.ThefirstgroupincludesdailyvaluesofT, samefourmeteorologicalvariablescharacterisingtheweatherof
Plantsrememberpastweather 185
antecedentdays,aswellasoneresultantvariableincludingactual pollen types; the longest pollen season during the 11-year
dailypollenconcentrationsofthegiventaxa)inordertotrans- period was considered for each year (see Section 2.1).]
form the original variables into fewer uncorrelated variables. Furthermore, antecedent meteorological conditions in the
These new variables, called factors, can be viewed as latent pollen-free season preceding the actual pollen season were
variablesexplainingthejointbehaviourofthepastandcurrent also considered. First take, the matrix comprising the pollen
meteorological elements as well as the current pollen and meteorological data of the 1st day of the pollen season
concentration. (April 16) for the 10-year period 1998–2007, as well as the
Afterperformingthefactoranalysis,aspecialtransforma- meteorologicaldataofthelastdayoftheprecedingpollen-free
tionoftheretainedfactorswasperformedtofindouttowhat season. In this way, we receive a matrix with ten lines (ten
degree the above-mentioned explanatory variables affect the identicaldays,i.e.April16,forthe10years,1998–2007)and
resultant variable (daily pollen concentration) and to give a nine columns (a column for the daily pollen concentrations,
rank of their influence (Fischer and Roppert 1965; Jahn and four columns for the current meteorological data, i.e. daily
Vahle 1968; Jolliffe 1993). In more detail, when performing values of T, RH, GSR and P, as well as four columns for the
factor analysis on the standardised variables, factor loadings antecedentmeteorologicaldata,i.e.dailyvaluesofT,RH,GSR
received are correlation coefficients between the factors and andP,forthelastdayoftheprecedingpollen-freeseason).Inthe
the standardised original variables and, after rotation, the nextsteps,onlythefourcolumnsfortheantecedentmeteorolog-
coordinatevaluesbelongingtotheturnedaxes(namely,factor icaldatawillchange;namely,theywillbethecumulateddaily
values). Consequently, if the resultant variable is strongly valuesofT,RH,GSRandPforthelast2,3,…,185daysofthe
correlated with a factor (the factor has a high factor loading preceding pollen-free season, respectively. In this way, we re-
at the place of the resultant variable) and an influencing ceive 185 matrices with ten lines and nine columns. In the
variable is highly correlated with the same factor, then the followingstep,weconsiderthedatamatrixcomprisingthepollen
influencingvariableisalsohighlycorrelatedwiththeresultant andmeteorologicaldataofthe2nddayofthepollenseasoninthe
variable. Accordingly, it is advisable to combine all the first five columns, and then we follow the same reasoning as
weights of the factors, together with the resultant variable, above.Thereby,wereceiveanother185matriceswithtenlines
intoanewfactor.Namely,itiseffectivetorotatesothatonly andninecolumns.Repeatingthisprocedure180times,atlastwe
one factorhas great loadwith the resultant variable, and the receiveadatamatrixcomprisingthepollenandmeteorological
remainingfactorsareuncorrelatedwiththeresultantvariable, data of the 180th day of the pollen season in the first five
thatistosay,areof0weights.Thislatterprocedureiscalled columns, while the last four columns consist of the antecedent
special transformation (Fischer and Roppert 1965; Jahn and meteorologicaldata,i.e.cumulateddailyvaluesofT,RH,GSR
Vahle1968). andP,forthelast185daysoftheprecedingpollen-freeseason.In
Factor analysis with special transformation is used in the thisway,wereceivealtogether180×185=33,300datamatrices
study as follows. The first five variables of the above- forPoaceae.
mentionedninevariablesincludedailypollenconcentrations TheprocedureisthesameforbothAmbrosiaandPopulus.
anddailyvaluesofT,RH,GSRandPduringthepollenseason Altogether, there were 93×272=25,296 data matrices re-
ofagiventaxonforeveryyear.Theremainingfourvariables ceivedforAmbrosia,while53×312=16,536forPopulus.
(cumulativeT,RH,GSRandP)aredefinedasinSection2.2, All statistical computations were performed with
and factoranalysiswith special transformationis carried out MATLABsoftware.
for every day of the pollen season with every accumulation Notethatsomeofthestatisticalresultsweretransferredto
lengthin the pastfor eachtaxon,asinthe caseofthe linear thefigures(Figs.2,3,4and5).
regressionprocedureofSection2.2.Inmoredetail,generating
the data matrices in order to perform the factor analysis for
eachtaxonisasfollows.
Thecalculationprocedureandmainresultsarepresentedin
detailforPoaceae,becausethisspecieshasthelongestpollen 3Resultsanddiscussion
seasoninHungaryandhasbigcontributiontothetotalannual
pollenproduction.Inordertoassesstheeffectoftheanteced- Multiplecorrelationsofdaily pollencountsofthe threetaxa
ent and current meteorological conditions on the current onthedailyvaluesofthemeteorologicalvariablesaredeter-
Poaceae pollen concentration, the 1st-day, 2nd-day, …, mined(Fig.2).Asonly11yearsareavailablefortheanalysis
180th-day values of both the pollen concentration, and the and hence the sample size is very small, for every day, the
four meteorological elements of the current pollen season estimatedmultiplecorrelationshavebigvariances.Inorderto
were taken. [The duration of the Poaceae pollen season in removethisbigvariability,polynomialsasafunctionofdays
SzegedlastsfromApril16untilOctober12,namely180days arefittedtothedailymultiplecorrelationvalues.Theoptimal
(Table 1). Namely, both for Poaceae and the remaining two order of polynomials is determined by Akaike’s criterion
186 I.Matyasovszkyetal.
Fig.2 Multiplecorrelationsof
dailypollencountson
simultaneousvaluesofdaily
meteorologicalvariablesasa
functionofthedayoftheyear
Ambrosia Poaceae
Populus
(Akaike1974),andthesignificanceoftvaluescorresponding for Poaceae can bebestapproximatedbya third-order poly-
totheestimatedcoefficientsinpolynomialsischecked. nomial (at a significance level of 1 %) showing the highest
Itcanbefoundthatmultiplecorrelationsarequitemodest multiplecorrelationsatthebeginningofthepollenseason.A
anddonotnecessarilyexhibitannualcyclesduringthepollen local minimum and maximum can be observed around day
seasons of the three taxa (Fig. 2). The correlation has an 159 and day 230, respectively. The lowest multiple correla-
inclinationfordecreasefromthebeginningtotheendofthe tions occuratthe end ofthe pollenseason.Furthermore,the
pollen season for Ambrosia (dashed line), which is however correlations for Populus provide a clear decreasing trend
not statistically significant at reasonable significance levels (statisticallysignificantata3.5%level)(Fig.2).
(somewhathigherthan10%)basedonttest,sothatcorrela- The difference between the multiple correlations (factor
tionsshouldbeconsideredconstant(solidline).Correlations loadings)ofdailypollencountsonthecurrentvaluesofdaily
Fig.3 Differencebetweenboth
themultiplecorrelationsand
factorloadingsofdailypollen
countsoncurrentvaluesofdaily
meteorologicalvariablesandon
cumulativevaluesofdaily
meteorologicalvariablesasa
functionofthedayoftheyear
(horizontalaxis)andofthe
cumulationlength(verticalaxis),
Ambrosia
Ambrosia,multiple correlations Ambrosia,factor loadings
Plantsrememberpastweather 187
Fig.4 Differencebetweenboth
themultiplecorrelationsand
factorloadingsofdailypollen
countsoncurrentvaluesofdaily
meteorologicalvariablesandon
cumulativevaluesofdaily
meteorologicalvariablesasa
functionofthedayoftheyear
(horizontalaxis)andofthe
cumulationlength(verticalaxis),
Poaceae
Poaceae, multiple correlations Poaceae, factor loadings
meteorologicalvariablesandonthecumulativevaluesofdaily current and pastmeteorologicalconditionsoncurrent pollen
meteorologicalvariables asa function ofthe day ofthe year concentrationsfordifferenttaxahasmajorimportance.
andoftheaccumulationlengthiscalculatedandpresentedin Multiple correlations of daily pollen counts on simulta-
Figs.3,4and5.Iftheabovedifferenceispositive(negative), neous values of daily meteorological variables have no sig-
it indicates that the influence of the current meteorological nificant dependence on the day of the year for Ambrosia.
variables on daily pollen counts is higher (lower) than the However, a clear decreasing tendency in the correlations for
influenceofthepastmeteorologicalvariables. Populus can be explained as follows. The florescence of
ForAmbrosia,altogether93(numberofdaysofthepollen Populusstartsearliestcomparedtothatoftheremainingtwo
season)×272 (number of days from the first pollen-free day taxa. In the late February—early March, the start of flores-
following the previous pollen season to the last pollen-free cence strongly depends on the values of the meteorological
dayprecedingtheactualpollenseason)=25,296factoranaly- elements, basically on temperature. At the same time, in the
ses,whileforPoaceaeandPopulus,33,300and16,536factor latepollenseason,thehere-mentioneddependencegradually
analyseswithspecialtransformationswerecarriedout,respec- weakens that is confirmed by the absence of second flores-
tively (Table 1). Namely, in total, 75,132 procedures were cence(Király2009).
performed for every day ofthe pollenseason with (1) every CorrelationsforPoaceaeareapproximatedbyathird-order
accumulation length for each taxon in the past, and another polynomial of the day of the year. The phyto-physiological
75,132procedureswerecarriedoutforeverydayofthepollen background of this association can be explained as follows.
seasonwith(2)everyindividualdayforeachtaxoninthepast. Thereasonofthehighmultiplecorrelationsatthebeginning
Hence,altogether,150,264factoranalyseswithspecialtrans- ofthepollenseasoninearlyAprilisduetothefill-upofthe
formations were performed. However, due to the smaller waterstorageofthegroundwaterasaresultofsnowmelting
variabilityoftheassociations,onlythosefor(1)areanalysed andspringtimeprecipitation,increasingdailymeantempera-
inthestudy(Fig.3). tures, ceasing of night frosts and lengthening of daytimes
Pastweathermayplayarelevantroleinthecurrentdevel- (Láng 1998; Haraszty 2004). The here-mentioned compo-
opmentofplantsandtheirphenologicalphases.Forinstance, nents favour both photosynthetic and generative processes
afterextreme dry (wet) summers oryears,pollenproduction (e.g. pollenproduction).Due tothe continental rainfallpeak
ofdifferenttaxamay decrease(increase)substantially (Láng in summer time, pollen grains are washed out of the air;
1998; Haraszty 2004). Accordingly, studying the effects of furthermore, owing to the higher cloud coverage, plants
Fig.5 Differencebetweenboth
themultiplecorrelationsand
factorloadingsofdailypollen
countsoncurrentvaluesofdaily
meteorologicalvariablesandon
cumulativevaluesofdaily
meteorologicalvariablesasa
functionofthedayoftheyear
(horizontalaxis)andofthe
cumulationlength(verticalaxis),
Populus
Populus,multiple correlations Populus,factor loadings
188 I.Matyasovszkyetal.
receive smaller amount of sunshine involving a decrease in Concerning Poaceae, based on the two methods, strong
photosynthesis that affects generative processes, as well. In simultaneous higher weight of the current meteorological
this way, pollen production decreases in the first decade of conditions (positive differences with dark shade) occurs on
June (Makra et al. 2011). Following the continental rainfall the following days: 109, 132,160,181,208,230,240,249,
peak, the groundwater is restored and the cloud coverage 253and279.Ontheotherhand,theimportantroleofthepast
becomes lower again activating the metabolic processes. In meteorological conditions (negative differences with light
May,thefirstmowingofgrasseshasbeenoccurred,sodueto shade) is concurrent on days 153, 191–198, 210–212 and
thecontinentalrainfallpeak,grassesarecapableofregrowing 247. Flowering and pollen production of Poaceae cover the
again. However, in view of the higher temperatures, evapo- wholegrowingseasonthatisfarmuchlongerthanthoseofthe
transpirationintensifies,butplantsclosetheirstomasinorder remaining taxa. Furthermore, Poaceae responds to robust
to lose less water (Makra et al. 2011). Parallel to this, the weather events much more intensively and sensitively
degreeofphotosynthesisisdecreasingastheycannotassim- (Király 2009). Rainfall due to both weather fronts and local
ilateasmuchCO asbefore.Inthisway,themaximumpollen showerscontributestothedevelopmentofplantsand,inthis
2
production cannot reach its springtime level because less way, to the increase of their pollen production. During dry
energy and organic materials can be spent to pollen produc- periods, the groundwater ensures water supply for plants
tion.Severalgrassspeciesmovetothestateofaconstraintrest indicatingthehigherroleofthepastclimateelementsinthese
duringthesummer(withwithering)thatismanifestedinthe cases. The intensively stripped pattern, namely the frequent
decrease of the late summer pollen counts. This process is successionofthecurrent(past)climateelementsindicatedby
continued in autumn as both withered and secondly mowed dark (light) shades shows that Poaceae is the most water
grasses and grazed grasslands are unable to produce pollen sensitiveofallthetaxaanalysed(Horváthetal.1995)(Fig.4).
anymore. During the heat wave over Europe in the summer 2003,
Thedifferencebetweenboththemultiplecorrelationsand withitsmeantemperatureexceedingthe1961–1990meanby
factorloadingsofdailypollencountsonthecurrentvaluesof about5°CinJune,JulyandAugust,thegrasspollenseasonin
dailymeteorologicalvariablesandonthecumulativevaluesof Switzerlandwasmostaffectedstarting1–2weeksearlierand
dailymeteorologicalvariablesasafunctionofthedayofthe ending 7–33 days earlier than in general. Furthermore, ex-
yearand ofthe accumulationlengthispresented indifferent treme high Poaceae daily pollen concentrations were mea-
shadows of grey (Fig. 3). Results received using the two sured in Switzerlandin thatpollen season (Gehrig 2006). In
methodsrevealedcharacteristicsimilarities.Namely,extreme Hungary, due to a warming and drying climate, total annual
differences of the effects of the current and past weather pollenamountandannualpeakpollenconcentrationindicatea
elements featured by the different shades of grey indicate decrease for Poaceae. At the same time, based on the daily
pronounced coincidences. This assumes that there are real pollen counts for Poaceae, the warmest and wettest years
associations between phyto-physiological processes on one favour higher pollen release (Makra et al. 2012). In the
hand and the current and past meteorological elements for CarpathianBasininvolvingHungary,anextremelylargenum-
givenpartsoftheyearandgivenaccumulationlengthsofdays berofPoaceaespeciesoccurwithdifferentstartingandending
ontheother. datesanddurationoftheirlifecycles.Consequently,theeffect
ForAmbrosia,theimportanceofthecurrentmeteorological of the meteorological elements to the pollen production is
conditionsisclearlydetectedbybothmethodsondays198– differentforthedifferentPoaceaespecies.Ingeneral,amod-
200and243,whilesubstantialconcurrentnegativedifferences erate increase in temperature and precipitation is favourable
withthemajoreffectofthepastmeteorologicalconditionsare forthepollenproduction of the spring species, while they
emphasised by both methods on days 220, 263 and 278, have a smaller influence on summer species (Makra
thoughmultiplecorrelationsshowonlyaweakandflustered et al. 2012).
character of the difference on the latter 2 days in the year ConsideringPopulus,themostimportantroleofthecurrent
(Fig. 3). For the growth of Ambrosia, the appearance of the meteorologicalconditionswithbothmethodscanbedetected
continentalrainfallpeakand its degreeare determining.The concurrentlyondays79–80andday70butinthislattercase
duration, shifts and degree of this rainy period can substan- only for the accumulation lengths of 1–100 days back from
tiallydifferyearbyyear.This phenomenon, its occurrence thepresent.Atthesametime,thehighestconcurrentweightof
or absence, influences remarkably early stages of the thepastmeteorologicalconditionsoccursondays66–68,73–
growth and florescence of the taxon. Additional local 74and92,respectively.ForPopulus,atthebeginningofthe
showers in the summer can strengthen the effect of the frost-free period, predominance of the effect of the current
current meteorological elements. In warm and dry pe- climate elements is a characteristic for both methods. This
riods, the effect of the past climate elements becomes period is the start of the growing season in Hungary. Early
stronger that is manifested in former precipitation stored springtime temperatures and precipitation conditions are de-
in the soil (Stefanovits 1999) (Fig. 3). termining for the development of the plants. Due to the
Plantsrememberpastweather 189
precipitation,abigamountofwatercanbestoredinthesoil processes. Since they are common plants, the lack of
thatcontributestotheincreaseoftheeffectofthepastclimate propagulum does not limit their fast recovery. The recovery
elements during dry periods, namely in the first and second itself can happen several times even within the year, but
decadesofMarch,aswellasatthebeginningandthesecond landscape-level local natural or human disturbances do not
half of April. The role of the current climate elements is limitthese species asthey haveanextendedappearanceand
stronglydominantinthethirddecadeofMarchindicatingan distributioninthelandscape.However,forsomespecies(e.g.
increased role of temperature and precipitation. At the same forAmbrosia),humandisturbances(e.g.land-usechanges)are
time, predominance of the effect of the current climate ele- favouredandevenrequiredfortheappearanceandspread.We
ments isa muchlesscharacteristicinthe first halfofMarch stressselectingAmbrosia,sincelargeareasareusedasarable
that can be associated with late winter frosts or early begin- lands inthe Szegedareaand the lackofproperlatesummer
ningofspringtime(Láng1998;Haraszty2004)(Fig.5). soil works can enhance their presence and spread several
Meantemperature,relativehumidity,globalsolarradiation times,especiallyonsandsoils.Ifanarablelandisabandoned,
andprecipitationamountasthemostimportantmeteorologi- Ambrosiaissqueezedoutfastbynaturalcompetitorsofweeds
cal factors influencing pollen production of the taxa consid- andplantsformingnaturalassociations.Inthisway,adistur-
ered are confirmed by other authors (Bartková-Ščevková bance pattern and its degree are permanent for the analysed
2003; Kasprzyk and Walanus 2010; Ščevková et al. 2010; landscape. Herbs, especially Ambrosia, are adapted to these
Sabariegoetal.2011;Aboulaichetal.2013).InHungary,due disturbances.Notethatthesensitivityofherbstodisturbances
toawarminganddryingclimate,pollencountcharacteristics is different, and modelling disturbances as influencing vari-
(total annual pollen amount and annual peak pollen concen- ablesisverydifficult.Annualtrendofragweedpollenlevels
tration)indicateadecreaseforAmbrosia,whileforPopulus, inthefunctionofthechangeinlanduseashumandisturbance
an increase is expected (Makra et al. 2012). Concerning cannotbedetermined.Ontheotherhand,changeinlanduse
extreme weather events, the coldest and the wettest years fortheSzegedareawasnegligiblefromtheyear1990tothe
highly facilitate Ambrosia pollen production, while both for year2006usingCORINELandCoverDatabase(http://www.
Poaceae and Populus, the warmest and wettest years favour eea.europa.eu/publications/COR0-landcover). Accordingly,
higher pollen release (Makra et al. 2012). Furthermore, land-use changes do not influence the pollen concentration
Matyasovszkyetal.(2012)establishedatendencytostronger of Ambrosia over the Szeged area in the period examined
associations between the meteorological variables and the (Deáketal.2013).Contrarily,extremevaluesoftheclimatic
ragweed pollenvariablein the pre-peakpollenseason, com- parametersasnaturaldisturbancesdoaffectdailypollencon-
pared to the post-peak pollen season (late summer to early centrations, and since their daily values are available, their
autumn) that is confirmed by Giner et al. (1999) and Laaidi currentandpasteffectstothetargetvariablecanbemodelled.
(2001). This is in accordance with the statement of Makra Accordingly,theanalysiswasperformedfortheabovetwo
et al. (2014) who found that definitely stronger associations herbaceoustaxaaswell,sinceitservedaclearpossibilityfor
betweenthemeteorologicalvariablesandthePoaceaepollen comparingthedifferenceintheweightofboththecurrentand
variablearefoundinthedatasetsforlowersummertemper- past meteorological conditions on the pollen production of
atures with near-optimum phyto-physiological processes both herbaceous and arboreal taxa. Note that diurnal pollen
compared to those for high and extreme high temperatures concentrationsoftreesaremuchmoreirregularwithouthav-
modifying life functions and, accordingly, interrelationships ingaclearandrecurrentannualpatternthanthoseforherba-
ofthemeteorologicalandpollenvariables.Thisisduetothe ceous taxa (Kapyla 1984; Galán et al. 1998). Furthermore,
differenceinthebehaviouroftheplantstostandenvironmen- onceanthesishasstartedintrees,itisrelativelyindependentof
tal stress. The above findings are clearly confirmed by the weather variables (Kapyla 1984). In accordance with the
decreasing multiple correlations of daily pollen counts on abovestatement,higherclimatesensitivityoftheherbaceous
simultaneousvaluesofdailymeteorologicalvariablestowards taxa can be definitely established compared to that of the
theendofthepollenseasonforallthethreetaxa. arborealPopulus.Namely,thefrequentsuccessionofthedark
Herbaceousplants(forourcaseAmbrosiaandPoaceae)are (white)stripsindicatingthepredominanceoftheeffectofthe
generally more sensitive to disturbances; however, they re- current (past) climate elements shows an obviously higher
cover well faster compared to arboreal plants (for our case climatesensitivity(especiallywatersensitivity)forAmbrosia
Populus) (Jones and Harrison 2004). Therefore, one may andPoaceaethanforPopulus(Figs.3,4and5).
conclude that studying past meteorological conditions on Note that both multiple correlation analysis and factor
thesespeciesisunnecessary.However,therecoveryofthese analysiswithspecialtransformationcanonlybeappliedifit
herbs from natural disturbances (e.g. floods, drought and is assumed that the relationships between the variables are
extremevaluesoftheclimaticparameters)andhumandistur- linear.Theassociationsanalysedinthestudycanbenonlinear;
bances(e.g.land-usechanges)isanessentialfactor.Seedbank hence,ourresultsarepossiblydistorted.However,ourmeth-
and nearby propagulum sources help the regeneration odologies are capable of separating the joint effect of the
190 I.Matyasovszkyetal.
current and past weather elements respectively, and, in this lastwesolvedthetasksuccessfully.Thesefigures(Figs.3,4
way, the results received can be considered as first steps and 5) (1) show specific and unique ‘imprints’ of the taxa
towardsdiscoveringthesenonlinearrelationships. considered; namely, they are characteristics to each taxon,
We should remark that only 11-year data sets were avail- and, in turn, each taxon has its own ‘imprint’; (2) the here-
ablethatinvolvealimitationfortheanalysis.Thedifference mentioned‘imprint’ofagiventaxonreceivedusingthetwo
betweenboththemultiplecorrelationsandfactorloadingsof methodologiesaresurprisinglysimilar;(3)wecaneasilyfind
dailypollencountsonthecurrentvaluesofdailymeteorolog- that arboreal Populus shows the widest, while herbaceous
icalvariablesandonthecumulativevaluesofdailymeteoro- Ambrosia and Poaceae, especially Poaceae, show the most
logicalvariablesasafunctionofthedayoftheyearandofthe narrowstripsoflightanddarkshadows.Theselatterfindings
accumulation length was based on only 11 data on identical maydenotethedifferentsensitivityofthesetaxainreturnto
days of the year corresponding to the 11-year data set. their reaction with the changing meteorological parameters.
Therefore,thereisnosensetoevaluatethestatisticalsignifi- The here-mentioned findings confirm our results and ensure
cance ofthe differencesamong the correlationsindicated by us that the methodologies applied highlight well the most
the differentshadows ofgrey. Hence,theshort data setcon- importantcharacteristicsofthetaxon-specific‘imprints’.
strainsnottoinvolvefurtherinfluencingvariables. Thequestionarises,whethertheresultsreceivedarevalid
Phyto-physiologicalassociationsoftheclimateparameters forotherlocationsaswell;namely,generalinferencescanbe
usedareasfollows.Globalsolarfluxhasamajordirecteffect drawnortheyarelocationdependent.Theinferencescannot
onthephotosyntheticprocesses,aswellasanindirecteffect be independent of the environmental conditions of a certain
ontheplants’developmentthroughthetemperatureparame- regionasthespeciescompositionandmetabolismfeaturesare
ters. The start and run of the photosynthetic and generative adaptedtoacertainclimatezone.Ofcourse,forareashaving
processesoccurinalimitedrangeoftemperature.Inthisway, similar species (at least genus), the composition can have
extreme global solar flux involves too high or too low tem- similaranswertothemeteorologicalvariablesinvolved;how-
peratures,and, due tothis, the available water may limit the ever,resultscanbeinterpretedjustforthetemperateclimate
vegetative and generative processes (seewinterand summer zone.
periods)(Makraetal.2011;Deáketal.2013).Concerningthe The analysis of the relationship of the past values of the
fourclimaticparameters,thecurrentandpastcomponentsof meteorological elements with the actual pollen season is an
temperature and precipitation indicate a major effect on the issue of great importance in aerobiology. Though there are
currentpollenconcentration.Long-timeeffectoftheprecipi- somestudiespresentingtheimpactofpastweatherconditions
tation is manifested in the storage of groundwater or in the on certain phenological phases of different taxa, only few
storedwaterintheplants.Relativehumidityisalargevariable papersdealwiththisarea.Laaidi(2001)establishedingeneral
parameter that has a rather small long-time effect on the thatheatandenoughrainfalltosatisfythewaterrequirements
current pollen production. An optimum relative humidity is duringthemonthbeforepollinationallowaplanttocomplete
necessaryfor theplantsfor performinggenerativeandvege- itsdevelopment,andinparticularitspollenproduction,andso
tative (e.g. photosynthesis) processes smoothly. Water cap- pollinate earlier. Spieksma et al. (1995) found that the air
tured in stomas is essential for these processes. However, temperature during the preceding 40 days has a decisive
optimum relative humidity assumes optimum temperature influence on the start date of the Betula pollen season. In
and precipitation conditions. Relative humidity has a major Melbourne, Ong et al. (1997) used only the rainfall sum of
role only in dry climate conditions (see the Mediterranean, July for developing a regression equation that predicted the
deserts, steppes, savannas), where the limitation or lack of beginning of grass pollination (which takes place between
groundwater is substituted partly by relative humidity. October 20 and November 24) with a satisfactory precision
However, these conditions are not typical in the temperate of76%.Kapyla(1984)andGalánetal.(1998)detectedthat
belt (Láng 1998; Haraszty Á 2004; Makra etal. 2011; Deák diurnalpollenconcentrationsoftreesweremuchmoreirreg-
etal.2013).Windspeedisomittedfromconsideration,dueto ularwithouthavingaclearandrecurrentannualpatternthan
its irrelevant role as a component of the past meteorological those of herbaceous taxa. Furthermore, once anthesis has
conditions. started in trees, it is relatively independent of weather vari-
The association of the pollen production with the climate ables (Kapyla 1984). Emberlin and Norris-Hill (1991)
elementsdiffersforthedifferentspeciesthatisclearlyrepre- established that annual differences in the cumulative
sentedinthestudy(Figs.2,3,4and5).Its reason is that the Urticaceaepollenconcentrationwereprimarilyduetoweather
different taxa show very different dispersal behaviour conditions in the period of pollen formation and only
with the influencing variables used in the study (Jones secondarily due to weather conditions in the pollen release
and Harrison 2004). season. Furthermore, relative humidity, temperature, wind
Forus,itwasareallybigchallengetofindawaytoshow speed and rainfall were most important in daily variations,
these associations for the readers.However, we think thatat buttheirrelativeimportancevariedbetweenyears.Gineretal.