Modeling radial growth increment of black alder (Alnus glutionsa (L.) Gaertn.) tree

  • Published on

  • View

  • Download


ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189avai lab le at iencedi rec t .comjourna l homepage: www.e lsev ier .com/ locate /eco lmodelModeling radial growth increment of black alder (AlnusglutioJana Lag akba Laboratory 3, Nb Departmen ubljaa r t i c l e i n f oArticle history:Published on line 8 April 2008Keywords:Wetland forBlack alderForest growFeature seleMachine leaSTELLA moda b s t r a c tNowadays it is extremely important to understand ecosystem function and its dynamics topredict future changes and consequently to perform appropriate measures. Hydromeliora-1. InThe undersof ecologicwhich we hand interesclimate contems is thagrowth is d CorresponTel.: +386 53E-mail a0304-3800/$doi:10.1016/estth modelctionrningeltions and subsequent decrease in groundwater table are thought to be a major reason fora decline in the vitality of black alder (Alnus glutinosa (L.) Gaertn.) wetland forests in North-eastern Slovenia. In this study radial increments of trees were used as indicators of blackalder forest function and its disturbances. The aim of the study was to build a model ofannual radial increments of black alder trees, to use this model to identify environmentalattributes that most importantly affect ecosystems function and to predict changes in theforest function under different scenarios of environmental conditions in the future. Themodel was induced with a machine learning algorithm CIPER and it was based on the dataabout site conditions and applied management measures in the past 35 years. Groundwa-ter levels in combination with the duration of sun radiation were identied as the mostimportant environmental attributes affecting annual radial increments. Radial incrementswere the lowest in very wet and cloudy years. On the other hand, radial increments weredecreased under drought stress aswell. Changes in groundwater level and in duration of sunradiation, aswell as increased oscillations of groundwater level, all cause important increasein oscillations of modeled radial increments, indicating higher stress. Radial incrementswere further negatively affected by late white frosts in the spring. 2008 Elsevier B.V. All rights reserved.troductiontanding of ecosystem function and reconstructional niche are especially important in this period inave to carefully balance between different needsts and in which we expect important changes ofditions. An important property of forest ecosys-t they mark annual radial growth increments. Asependent on ecosystems well-being and functionding author at: University of Nova Gorica, Laboratory for Environmental Research, Vipavska 13, SI-5001 Nova Gorica, Slovenia.3 15 328; fax: +386 533 15 296.ddresses: (J. Laganis), (A. Peckov), (M. Debeljak).in individual years, radial increments can be regarded as reli-able indicators of ecosystem function.Black alder (Alnus glutinosa (L.) Gaertn.; f. Betulaceae) is adeciduous tree species with many special ecological prop-erties as well as of an economical importance. Its mostimportant ecological properties are adaptations to highgroundwater level, nitrogen xation through symbiosis withactinomycetes Frankia, fast growth and short lifetime, andlight pretentiousness. In the stands under study it is also see front matter 2008 Elsevier B.V. All rights reserved.j.ecolmodel.2008.02.018nsa (L.) Gaertn.) treeanisa,, Aleksandar Peckovb, Marko Debeljfor Environmental Research, University of Nova Gorica, Vipavska 1t of Knowledge Technologies, Jozef Stefan Institute, Jamova 39, Ljova Gorica, Sloveniana, Sloveniaecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189 181reported to be resistant to white frosts, diseases and herbi-vores (Nemesszeghy, 1986; Brus, 2005).Homogenortheastenatural we(Nemesszeicantly redvitality ofmiddle of tLevanic, 19in Europe (to a decreachangedhyrations andand hydrolothe studiesently clearand standceed to profor forest de1993; Caterwe conductforest of PoRadial genvironmeparticular smate condgrowth of tbination ofreason we ction and disin individuThe maiulation growas used tothat affectand to predfuture clim2. Ma2.1. Stu2.1.1. SiteThe forestleft side ofin these stand oak (3%ble alder fo(Culiberg, 1best sites a30m are re1953; DawsFeatherstonstand undein individuBlack al(Forest Maalready suruary 2005. The latitude of the research site is 46,595 and itslongitude is 16,358. The area is at and it is about 190m aboveleveClilectemme. Imnt lain thd tolturaver, 19owinmmeties pr, 19umn thj, 19s is tmon ofandowlyIn ousurfanuas ad wieic ein deevanhysiantling deralen isHisingofW18thvereedsts wic anaininwatg (Le953 aThend ffor assze980erbenous forests of black alder in the lowlands ofrn Slovenia are among the last remnants oftland forests of this species in central Europeghy, 1986). The area of these forests was signif-uced in the 19th century. Severe decline in thestands in the studied area was observed in thehe last century (Wraber, 1951; Nemesszeghy, 1986;93) as well as in other black alder oodplain forestsPretzell et al., 1997). This decline was attributedse in groundwater table (Pretzell et al., 1997) anddrological conditions due to extensive hydromelio-regulations. The importance of groundwater levelgic regime for wetland trees was also conrmed inof Keeland and Sharitz (1997). Despite the appar-causal relationship between the groundwater levelvitality, studies performed until now did not suc-ve changes in groundwater level as a major reasoncline (Levanic andKotar, 1996; Kosir, 1987; Levanic,, 2002). In order to continue with this discussioned a case study research in a black alder oodplainlanski Log.rowth of trees is inuenced by a complex ofntal parameters (attributes) that take place on aite. These attributes include water availability, cli-itions and soil fertility (Whitehead, 1998). Radialrees responds very dynamically to a current com-environmental attributes (Waring, 1987). For thatonsidered radial increments as indicators of func-turbances of black alder wetland forest ecosystemal years.n goal of this study was to construct a reliable sim-wth model of black alder forest stand. This modelidentify the set of the most important attributesgrowth and development of the stand under studyict changes in the function of these stands underate change scenarios.terials and methodsdy sitedescriptionof Polanski Log covers the area of 414ha on thethe Ledava River. The most common tree speciesands is black alder (85%), followed by ash (12%)) (Nemesszeghy, 1986). Long-term presence of sta-rests was proved in a research of subfossil wood989). These stands can be classied as black aldersnd as a climax community. Whereas heights up toported for black alder in other countries (McVean,on and Funk, 1981; Brus, 2005; Krstinic et al., 2002;e, 2003), they surpassed the height of 34m in ther study (Nemesszeghy, 1986; Laganis, 2007). Treesal plots are even-aged.der represents 95% of trees in the selected standnagement Plans, 19712011). With 69 years itpassed its maturity and it was cut down in Jan-the sea2.1.2.The sehot su910 Cfrequefrostsreporte(SilvicuThe(Wrabethe grthe suproper(Wrabemaximbetwee(SmoleRiver aThefunctiowatervery sl1993).to the2004Jown.Thicoverehypoglvaries1951; Lpoor pAscendprevaiof minNitrog1988).2.1.3.Accordreportsin thewas coperformof fore(Levanof remgroundoodinyears 160 cm.2005) apriate(NemeIn 1the rivl.mate and soil conditionsd site lies in a Panonic-type of climatewith dry andrs and cold winters. Mean annual temperature isportant for the vegetation are negative impacts ofte frosts in the spring (until April) and early whitee autumn (Wraber, 1951), despite they were notcause important damage in black alder standsal chronicles).rage amount of precipitation is only about 800mm51), a bit less than 60% of which occurs duringg period (Ziberna, 1992). Severe droughts duringr and relatively poor physical and chemical soilrevent higher agricultural productivity in this area51). The amplitude between the minimum andgroundwater level was found to be about 1.5me years 1953 and 1993 in about 7-km distant well95), which is of a similar distance from the Ledavahe selected important attributes that affect structure andthese forests are hydrological conditions of soilwater regime. Soil water is standstill or it owsthrough a gravely substrate (Wraber, 1951; Levanic,r stand the groundwater level was relatively closeace (080 cm) during measurements (Septemberry 2005). In some cases the stand was partly over-t area has gravelly to sandy siliceous grounding,th fertile, clayey alluvium (intrazonal type of soil,ugley; Wraber, 1951; Kalan, 1988). This alluviumpth but in general it is relatively shallow (Wraber,ic, 1993). A shortage of limestone is a reason forcal and chemical soil properties (Wraber, 1951).groundwater ows toward the surface (upward) areuring the vegetation period and they bring plentynutrients to the upper soil layers (Kalan, 1988).reported to be abundant (Levanic, 1993; Kalan,toryto the Forest Management Plans (19591968) andraber (1951) large part of the area was uncrossablecentury due to high water levels. The landscaped by oak and alder forests. Several meliorationsafter the year 1814 drained the area, large extentsere cleared to obtain new agricultural surfacesd Kotar, 1996; Nemesszeghy, 1986) and large partg forests was endangered due to a decrease in theer level (Wraber, 1951) and changes in the rhythmofvanic, 1993). Smolej (1995) reports that between thend 1992 (40 years) groundwater level decreased forarea became less appropriate for black alder (Brus,orests remained only on soil, which were inappro-griculture due to high groundwater or low fertilityghy, 1986).a reservoir at Radmozanci was constructed andd of the Ledava River was deepened. As a result a182 ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189further decrease of groundwater level took place, the naturalooding in the springwas halted (Kalan, 1988; Cater, 2002) andan articial ooding in the summer was introduced (Levanic,1993; Levanic and Kotar, 1996).2.2. Methods2.2.1. DatasetDataset (attributes) required for the model construction andvalidation consisted of the data on the radial growth of eighttrees in the selected forest stand, attributes describing mete-orological and hydrological conditions and attributes on theforest management measures performed in the selected for-est stand. All three types of attributes were available for thepast 35 years before the nal clear-cut of the selected stand.During the felling we obtained tree disks from eight neigh-boring trees from the dominant social layer. Neighboring treeswere chosen to minimize differences in site conditions andmanagement measures. About 5-cm-thick tree disks weretaken at the trunk height of 1.3m. After air-drying, sand-ing and polishing they were prepared for dendrochronologicalanalysis. Radial increments were measured to the nearest0.01mm using a LINTAB measuring stage and a dissectingstereomicroscope Olympus SZ-CTV (SZ-60) with video display.Measurements were done in two different directions on eachtree disk and the average value was used in the calculations.Meteorological attributes were obtained from the nearestpermanent meteorological station. As the area is at the datafrom about 5km distant Meteorological Station Lendava wereconsidered to be valid for our stand. Among meteorologi-cal attributes we investigated monthly data on duration ofsun radiation (h), precipitation (mm), potential evapotranspi-ration (ETo; mm), the difference between precipitation andETo, number of days with white frost, number of days withsnow, maximum, average, and minimum monthly temper-atures (C), cumulative temperatures above 0 C, above 5 C,and above 10 C, number of days with minimum temperatureabove 0 C, below 4 C, below 10 C and above 25 C, as wellas number of days with maximum temperature above 10 Cand above 25 C.Groundwater levels were measured in about 7km dis-tant well and the Ledava River levels were measured on apermanent measurement station Centiba, which is locatedabout 10km downstream from the research site. We acquiredattributes on monthly minimum, average, and maximumLedava River levels and groundwater levels. Our stand islocatedwithin the LedavaRivers catchment area and theRiverows a bit more than 1km away from our stand. Previousresearch conrmed a close link-up and synchronous uctu-ations of Ledava River level and groundwater level in this area(Levanic, 1993). Groundwater levels showed poor agreementwith radial increments (e.g. Table 1) and they were not used inthe further work. Poor regression coefcients most probablyindicate that the sampling well was too far away and that itwas affected by some other hydrological regimes. At the sametime the data on this attributewere available for a shorter timeperiod (20 years).The attributes on forest management measures per-formed in the selected forest stand in the past 35 years(Forest Management Plans, 19711980, 19811990; 19922001;Table 1 ion cAge (0.6 erLthinn (0.3 un 4thinn y-1 ( un 4thinn y-2 ( un 4maxgw 7 (0 un 4maxgw 47 un 5maxgw 81 un 4maxgw 41 ecip 7mingw 7 (0 ecip 8mingw 47 ecip 4mingw 81 ecip 8mingw 41 ecip 4avergw 7 (0 ecip 4avergw 47 ec-ETavergw 81 ec-ETavergw 41 ec-ETmaxL 5 (0 ec-ETmaxL 7 (0 ec-ETmaxL 47 ( ec-ETmaxL 810 o 47maxL 410 o 81Numbers r to theFebruary o July);maximum son;duration of iratiand potent peratwith the m inimuwith whiteThe selected attributes for CIPER and corresponding regress19) maxL 57 (0.200) av98) maxL 59 (0.192) t-s0.149) minL 4 (0.533) t-s0.075) minL 5 (0.609) t-s.205) minL 7 (0.613) t-s(0.062) minL 8 (0.582) t-s0 (0.016) minL 9 (0.641) t-s0 (0.055) minL 47 (0.665) pr.177) minL 810 (-0.645) pr(0.082) minL 410 (0.684) pr0 (0.046) minL 46 (0.627) pr0 (0.062) minL 5-7 (-0.683) pr.194) minL 58 (0.690) pr(0.064) minL 49 (0.668) pr0 (0.017) minL 122 (0.548) pr0 (0.026) averL 5 (0.553) pr.282) averL 47 (0.600) pr.086) averL 80 (0.553) pr0.092) averL 410 (0.658) pr(0.150) averL 46 (0.613) ET(0.058) averL 49 (0.635) ETefer to the month(s) of the year the attribute refers to (e.g. 122 refersf the current year; 47 refers to the period between current April andvalues respectively; thinn: thinning (m3/stand) before the current seasun radiation (h); prec: precipitation (mm); ETo: potential evapotranspial evapotranspiration (mm); T: temperature; cumT>0: cumulative teminimum temperature below 0 C; d-minT>25: number of days with mfrost; d-snow y-1: number of days with snow in the previous winter.orrelation coefcients in the parenthesis122 (0.422) ETo 410 (0.072)(0.362) maxT 6 (0.297)7 (0.392) maxT 7 (0.031)10 (0.383) maxT 410 (0.073)6 (0.414) maxT 57 (0.130)9 (0.417) maxT 122 (0.049)9 (0.439) minT 4 (0.066)(0.233) minT 8 (0.126)(0.114) minT 47 (0.066)7 (0.155) minT 410 (0.118)10 (0.085) minT 46 (0.066)10 (0.166) minT 49 (0.066)9 (0.182) averT 47 (0.020)o 7 (0.226) averT 410 (0.078)o 8 (0.120) cumT>0 47 (0.005)o 47 (0.155) cumT>0 410 (0.073)o 810 (0.092) d-minT25 6 (0.147)o 49 (0.177) d-wf 47 (0.399)(0.054) d-wf 410 (0.130)0 (0.103) d-snow y-1 (0.141)period between December of the previous year and themax, aver and min refer to the minimum, average andgw: groundwater level; L: the Ledava River level; t-sun:on (mm); prec-ETo: the difference between precipitationure above 0 C in this period; d-minTecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189 183Table 2 A comparison of relative root mean square error (RRMSE) and indicators of the equation complexity of 12models chosen after the rst selectionM n sizea Equation degreea Equation lengtha1 Jn 22 Jn 83 Jn 84 Jn 75 Jn 16 Jl 17 Jl 28 A 59 Jn 110 Jn 711 A 412 Jn 6a The size uatiolength is20022011)mercial thinand beforecommonlyon biomassAccordinAntonic etRepublic ofthat the grber. We invperiod as wor extremeand July (5and Auguswhich were2.2.2. DaTo identifyradial growannual radment attribapplied.CIPER sEquationsalgorithm f(Todorovskthrough theisfy the givpolynomialand ts theIn the salways nddata. Howeoptimal tratting the(MDL) Princmethod fortion to theCIPER benomial. Thmultiplyingon cace oce oon.IPERsamdataaltoghad tIPERincreagemibuteseleFiftydurinnstrid inMost sethe rodel RRMSE Number of attributesin the equationEquatioj3 2m 0.7282 6 1j3 3s 0.7599 6j3 1s 0.7614 6j3 4m 0.76455 3j2 2 0.7685 5 1y 4xl 0.7686 6 1y 3al 0.772 6 1vg 3hxl 0.7746 6j2 3m 0.7748 6 1j1 4 0.7839 3vg 3al 0.7887 4j2 2s 0.8060 5of the equation is the number of polynomial terms in the equation; eqthe sum of the exponents of the variables.consisted of themass ofwood removed by precom-ningperformedbefore the current growing seasongrowing seasons 1, 2 and 3 years ago (thinning isrealized in thewinter). Available were also the dataof wood and biomass increments in each decade.g to the results of Chmielewski and Rotzer (2001),al. (1998/99) and the data from the Agency of theSlovenia for Environment (ARSO) we determinedowing season extends between April and Octo-estigated attributes for each month during thisell as attributes aggregated into groups: averagedvalues between April and June (46), between May7), between July and August (68), between Mayt (58), etc. Altogether we obtained 333 attributes,tested for their effect on radial increments.ta analysisthe attributes that most importantly affectth and to determine the relationship betweenial increments and environmental and manage-utes, the machine learning algorithm CIPER wastands for Constrained Induction of PolynomialequatiformaninuenequatiIn Cat theinitialyears,rstlywith Cradialor manof attrTherithm.testedent coresulte2.2.3.The ring tofor Regression (Peckov et al., 2006). This is anor inducing polynomial equations from the datai et al., 2004). The algorithm heuristically searchesspace of possible equations for solutions that sat-en constraints. The output of CIPER consists of theequation that satises the complexity constraintsdata best (Peckov et al., 2006).pace of polynomials of arbitrary degree we cana polynomial with error of zero on the trainingver, such equations can be very complex. To nd ande-off between complexity of the model and welldata the CIPER uses Minimal Description Lengthiple as a search heuristic. The MDL Principle is ainductive inference that provides a generic solu-model selection problem (Peckov et al., 2006).am search procedure starts with a constant poly-e equation is then rened with adding terms orterms with some attributes. Every term in thean indicatochosen amSTELLA sysSelected eqradial increing criteriochanging ecomplexitywell.Modelsues in indiradial incre(e.g. Figs. 2sitivity anathe modelevalues of ingiven percewere chang2 193 133 123 133 193 202 107 143 196 223 73 10n degree is the largest degree of any one term; equationn be considered as a factor that inuences the per-f the model. Only attributes that have signicantn the performance of themodel are included in thethe number of attributes that can be considerede time is limited according to the number of thewehave (in our case eight trees and the period of 35ether 280 data for each attribute). Consequentlyweo reduce the number of attributes for further work. We applied linear regression on the data aboutments of eight trees and individual environmentalent attributes. This way we reduced the numbers from 333 to 84 (Table 1).cted 84 attributes were treated with CIPER algo--twodifferent combinations of 610 attributeswereg the study. Each combinationwas runwith differ-ctions on the model complexity, which altogether124 experiments.delinglection among the models was performed accord-elative root mean square error (RRMSE), which isr of the error of predictions. Twelve models wereong 124 experiments for a more precise study intems thinking software (Isee Systems) (Table 2).uations were used to construct growth models ofments of black alder trees under study. The select-n applied to thesemodels was their behavior undernvironmental conditions (data not shown). Their, especially the equation degree, was considered ashelped us to compare modeled and measured val-vidual years (e.g. Fig. 1) and changes in modeledments under changing environmental conditionsand 3).We run several simulations and several sen-lyses. We focused on determination of changes ind radial increments after increasing or decreasingdividual attributes or attribute combinations for ant (Fig. 3). In the presented graphs attribute valuesed for 25%.184 ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189Fig. 1 Comradial increFig. 2 MoincreasingminL 810(t-sun 47,very dry anare presenradial increin the sprinobserved n3. ReRegressionLedava Riving radialselected atLedava Rivand numbBlack aldertemperaturotranspiratwithin theThe oveest RRMSE ichosen forand reasontion degreethat provedminimumminL 47) aAugust to Oin the seconof sun radiof the growfrost in themodel in a form of polynomial equation wasncrem)anientsintetioncient.comin Fincron ofparison of modeled and measured values ofments (r-incr; mm).deled values of radial increments underlevels of the Ledava River (maxL 810, minL 47,) and decreasing duration of the sun radiationt-sun 810). The left side of the graph representsradial i(mmOurcoefcdencecalculafour smodelThesentedunderduratid sunny years, whereas wet and cloudy yearsted on the right. Three curves present modeledments at three different levels of late white frostg (d-wf 47): minimum, average and maximumumber of days with white frost.sultscorrelation coefcients showed that minimumer levels were the most important attribute affect-growth in this ecosystem (minL; Table 1). Othertributes (Table 1) were maximum and averageer levels (maxL, averL), duration of sun radiationer of days with white frost in the spring (d-wf).radial growth seemed to be relatively indifferent toe uctuations, precipitation (prec), potential evap-ion (ETo) and to the drought stress index (prec-ETo)observed range of values.rview of the 12 selected CIPERmodels with the low-s presented inTable 2. The experiment jnj3 2mwasfurtherwork according to the complexity, reliabilityableness. It proved the lowest mistake, low equa-(Table 2) and the most reliable behavior. Attributesto be the most important in this model are age,Ledava River levels in the rst (from April to July;nd in the second part of the growing season (fromctober; minL 810), maximum Ledava River levelsd part of the growing season (maxL 810), durationation in the rst (t-sun 47) and in the second parting season (t-sun 810), number of days with latespring (from April to July; d-wf 47). The resultingsensitivityThe corments and0.877. Thisbecause CIindividualIn the athe optimathan averagaverage dudrier condithe incremobserved innicantly anumerousOverviewlowest radienvironme Low Ledason, highthe growgrowingfrost in t High Ledout the gfrost.In the foor furtherquently groduration ofLedava Rivent= + 0.0511025526922 minL 8 10+ 0.0291795197998 maxL 8 10+ 0.017479975134 t-sun 4 7+0.0346935385853 t-sun 8 10+ 1.950606536E 05 t-sun 8 102+ 2.01014710248 d-wf 4 7+9.35586778387E 05 minL 4 7t-sun 4 7+ 0.000179339939732 minL 4 7t-sun 8 10+6.45688563611E 05 minL 8 10t-sun 8 10+3.06551434164E 05 maxL 8 10t-sun 4 7+0.00282485442386 t-sun 4 7 d-wf 47+ 0.00141078675225 t-sun 8 10d-wf 4 7+7.91071710872(1)alysis showed that the precision of individualdoes not importantly affect the outcome. Con-rvals were calculated for every coefcient. Thoses show that there is 95% condence that withic digits we are preserving the stability of theparison of modeled and measured values is pre-ig. 1. The graph on changes of modeled incrementseased levels of the Ledava River and decreasedsun radiation are shown in Fig. 2. Fig. 3 presentsanalyses.relation coefcient between modeled radial incre-the average measured increments of eight trees isagreement is higher than it is predicted in CIPERPER operates with the data on radial increments oftrees, not on their average values.bsence of white frost (r-incr. mind-wf 47 in Fig. 2)l conditions for the radial growth are at a bit belowe level of the Ledava River and at a bit above thanration of sun radiation (Fig. 2). In both wetter andtions (higher or lower levels of the Ledava River)ents decrease. Especially pronounced decrease ishumid and cloudy years. Frequent late frosts sig-ffect radial increments aswell. However, yearswithdays of frost were rare in the period under study.of the results reveals that the model predicts theal increments under the following combinations ofntal conditions:va River levels in the rst part of the growing sea-Ledava River levels through the second part ofing season, high sun radiation through the wholeseason, and more number of days with late whitehe spring.ava River and low sun radiation duration through-rowing season, higher number of days with whitellowing paragraphs we present effects of increaseddecreased levels of the Ledava River (and conse-undwater levels; Fig. 3A), increased or decreasedsun radiation (Fig. 3B), combinations of decreaseder levels and increased duration of sun radiationecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189 185Fig. 3 Sensitivity analyses: modeled radial increments under measured attribute values and after individual attribute orattribute combinations were increased or decreased for 25%. Increased and decreased were (A) all Ledava River levels:minimum levels between April and July (minL 47) and between August and October (minL 810), maximum levels betweenAugust and October (maxL 810); (B) duration of sun radiation (h) between April and July (t-sun 47) and between Augustand October (t-sun 810); (C) all Ledava River levels (minL 47, minL 810 and maxL 810) and opposite change in all sunradiation periods (t-sun 47, t-sun 810); (D) all minimum River levels (minL 47, minL 810) and opposite change inmaximum River levels and sun radiation periods (maxL 810, t-sun 47, t-sun 810); (E) number of days with white frostbetween April and July (d-wf 47).186 ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189).(drier and syears) (Fig.levels andsun radiatiwater levelyears witheffects offrosts in thChangesRiver (Fig. 3decrease ain sunny ysunny andThe effevalue of thduration ofdecrease inchange cauand/or wetresult in inThe effeels (minL 4and inverset-sun 810;affected byRiver levelsthe modelesun radiatiincrementsVery imRiver levelscause an inconditions.under condduration oincrease inincrementsa decreasedecreases dwet years.Except iber of daysradial increa decreaseDilectimaE) anluessed af the2).complecteincreith kcanspone reaAgsenclaineecteincrees (Teasesed ddesbe niablyFig. 3 (Continuedunnier years) and vice versa (wetter and cloudier3C), combinations of decreased minimum Riverincreased maximum River levels and duration ofon (drier and sunnier years with higher ground-uctuations) and vice versa (wetter and cloudierlower groundwater level uctuations) (Fig. 3D), andincreased or decreased abundance of late whitee spring (Fig. 3E).of increments under changed levels of the LedavaA) indicate that increments more or less linearlys the River levels increase. The response is weakears with plenty of precipitation, and stronger indry years.ct of sun radiation (Fig. 3B) is dependent on thee other attributes. Generally an increase in thesun radiation in otherwise sunny years causes aradial increments. On the other hand the sameses an increase in radial increments in cloudyyears. Both increase and decrease of sun radiationcreased oscillations of radial increments.ct of simultaneous change of the Ledava River lev-7, minL 810, maxL 810; e.g. we increased them)change in the duration of sun radiation (t-sun 47,e.g. we decreased them; Fig. 3C) is importantlythe appearance of white frosts and high Ledava(maxL 810). High River levels cause a decrease ind increments and they diminish the importance ofon levels. White frosts cause a decrease of modeledin the majority of cases.portant is also the effect of maximum Ledava4.The setion in(RRMSing vawere uness o(TableInthe seradialmentwmodelonly rewas th4.1.The abbe expwe expradialanalysa decrincreaThewouldTo rel(maxL 810; Fig. 3D). Increased maxL 810 levelscrease in radial increments in drier and cloudierIncreased maxL 810 levels have the same effectitions of reduced Ledava River levels and increasedf sun radiation. Moreover, the study showed thatmaxL 810 further increases differences in radialbetween wet and dry years. On the contraryin this attribute functions like a buffer and itifferences in radial increments between dry andn very dry and sunny years an increase in the num-with white frost (d-wf 47) causes a decrease ofments (Fig. 3E). In agreementwith our expectationsin the d-wf 47 stimulates radial on racorrespondmeteorolog4.2. ThradiationThe modelradial increconditionsIncreaseddifferencesing seasonin decreasescussionon of the best model remains an important ques-chine-learning tasks. In our work the reliabilityd the reasonableness of the model under chang-of individual attribute or attribute combinationss the most important indicator of the appropriate-model. Besides this, simple models were preferredarison to the other models under examinationd one shows much higher predicting power onments under study (Fig. 1) and much higher agree-nowledge-based expectations. The behavior of thebe explained through physiological processes. These, which cannot be regarded as wholly suitable,ction on frequent late white frosts in the of the attribute age in the selected model couldd by the maturity of the stand. For mature treesd slow and approximately linear negative trend ofments in time, as it was conrmed by statisticalable 1). On the other hand the model suggests thatof radial increments appeared due to the trend ofuration of sun radiation in the past decade.cription of changes in the radial growth in timeeeded to enable completely dynamic modeling.distinguish effects of age we would need thedial increments in some younger trees from theing stand or we would need longer series of hydro-ical and management data.e Ledava River level and duration of sunpredicts an important increase in oscillations ofments as soon as meteorological or hydrologicalimportantly moves from the observed conditions.oscillations of the Ledava River levels (increasedbetween the rst and the second parts of the grow-or between maximum and minimum levels) resultd radial increments and increased oscillations ofecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189 187radial increments, both changes indicating higher stress anddecreased stability of ecosystem function.Levanicsteadily deplots in theone plot thin the past don this ploto slowly aof them. Unlogical andrhizospherTrees thgroundwat1995; Brus,on the othebelow the w1956, 1953).the expectaKeelandthree oodwhereas trwater rootsthat long-laronment aconditionstem recons(Iremongerin agreemethat high grto these staThe imposcillationshigh waterooding (loon radial inhave negatgroundwatdrought takbe neededChangesradiation hments in dand D). Incin increasecauses a dewith low letions of theradial incresun radiatihigh grounthesis. Highoxygen intoSchroder, 1timehigherilates to thexation (Wand GodonPreviousKolenko, 2decrease in groundwater table was the major reason for adecrease in the vitality of these stands and consequently ase inat aappetenson threstrygen, 199rocerounveloal coeredgene(Nemgherconadialdec(andith ttreeultsundresuwithidenentswerouldwatgs gperimThng mowthitionller rentto dt inctedimrees,on ccondeigheasur ofbs eandt inced in9) inand Kotar (1996) found that radial increments werecreasing since 1965 in the majority of the researchneighboring black alder stands in Crni Log. Only oney observed a slight improvement of radial growthecade of the studied period. It is possible that treest, as well as the trees in this research, were abledapt to the new conditions and to take advantagefortunately we do not have the data on morpho-physiological changes of root system and in thee within this are growing on sites with permanently higher are known to develop shallow roots (Eschenbach,2005; Keeland and Sharitz, 1997). The roots of alder,r hand, are reported to have the ability to breakwellater table (Whyte and Sisam, 1949, cit. in McVean,This could be able to explain the trade-off betweentions and the modeling results in dryness.and Sharitz (1997) note that temporary ooding of-tolerant conifer species caused root deterioration,ees on permanently ooded locations developedwhich were adapted to inundation. They explainsting ooding results in a more stable soil envi-nd adapted root system. Trees growing in suchdo not have to invest a lot of energy into a root sys-truction and they show improved biomass growthand Kelly, 1988; Keeland and Sharitz, 1997). This isnt with the assertions of Levanic and Kotar (1996)oundwater uctuations represent themajor threatnds.ortance of groundwater levels and groundwateris stressed also in the results of this study. Bothlevels in otherwise dry seasons and absence ofw maxL 810) in wet years have positive effectscrements. High oscillations of groundwater levelive effect on radial growth. Increase in maximumer levels could be benecial only when seriouses place. More detailed physiological study wouldfor more accurate the Ledava River levels and duration of sunave the opposite effect on modeled radial incre-ry years compared to wet and sunny years (Fig. 3Crease in sunniness in otherwise wet years resultd radial increments, whereas the same changecrease in modeled increments in dry years (yearsvels of the Ledava River). Besides, these uctua-Ledava River level cause much higher changes ofments in cloudy years compared to yearswith highon. The model suggests that trees easily sustaindwater level in years with high rate of photosyn-photosynthesis could promote the transport ofthe roots (Iremonger and Kelly, 1988; Grosse and986; Schroder, 1989; Dilly et al., 2000). At the samephotosynthesis enables higher transport of assim-nodules and consequently higher rate of nitrogenheeler, 1971; Gordon and Wheeler, 1978; Dawson, 1979; Pizelle, 1984; Dilly et al., 2000).convictionwas (Nemesszeghy, 1986; Levanic, 1993;004; personal communication) that a moderatedecreagest thstresshigh inyears,tantlythe oxSharitzation phigh ghas dethe locto lownot toknownters higrowthalder rThelevelstrarywof thisthe reslow groThementany evincremmentsalder cgroundseedlinthe ex(1953).4.3.Thinnithe grof adda smainvestmas duedid nogenerathe lowadult tsivelyThe seclose nning mnumbestudy.HibgrowthA slighobserval., 198radial increments. On the contrary our results sug-decrease in radial increments due to the droughtars onlywhen both very low groundwater table andity of sun radiation takes place. Cloudy and rainye other hand, are declared to much more impor-ict radial increments. This could be explained bystress (Iremonger and Kelly, 1988; Keeland and7) and a simultaneous decrease in the nitrogen x-ss (McVean, 1953; Iremonger and Kelly, 1988) underdwater levels. These results suggest that the standped deep roots and that it is this way adapted tonditions. Consequently it is not endangered duegroundwater table. However, care must be takenralize these ndings to all stands in the area. It isesszeghy, 1986) that only some tenths of centime-altitude of a growing site can signicantly affectditions and signicantly change reactions of blackincrements on the groundwater level uctuations.rease of radial increments under high Ledava Riverconsequently high groundwater levels) is in con-he knowledge about the high degree of adaptationsspecies to high groundwater levels. In opposition towe expected a decline of radial increments underwater table (Levanic, 1993).lts obtained by the selected model are in agree-the ndings of Levanic (1993) who did not ndce that drought would cause a decrease in radialin these stands. He also found that radial incre-e often low in wet years. He concluded that blackgrow on these sites as long as it has access toer. A decrease in growth of black alder trees andrown in inundated soil was also conrmed inents of Kaelke and Dawson (2003) and McVeaninningeasures usually cause a subsequent increase inof the remaining trees as they gain advantageal resources. Only in the rst year after thinningeduction of growth can take place due to highers into the root system and into the canopy as wellisturbances of the soil system. In our case CIPERlude management-related attributes in any of theequations. Three possible explanations exist forportance of thinning in our stand. The rst is thatin contrast to young stands, do not respond inten-hanged conditions in the canopy (Levanic, 1993).possibility is that trees were not removed in theborhood of the trees under study during the thin-res. The third possible reason is a relatively lowthinning measures performed in the period undert al. (1989) found that thinning stimulated radialdepressed height growth of young red alder trees.rease of radial increments after thinning was alsothe study of Berntsen (1961, 1962; cit. in Hibbs et21- and 11-year-old stands.188 ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 1801894.4. White frostAccording tincrementstion intensat higher Rinsufcientwas the mothe modelThe mocannot be d4.5. TemLow imporblack alderof Levanicmean dailyply were imforest. Temradial growStravinskieOverallenvironmeThis seemeof trees us1997; Lebausharply aftsome growNovemberet al., 2003)can be proncontinuousand black aSeptember5. CoThe appliedtages and uof the selecthe responmental attrLedava Rivthe duratiodynamics odecrease inappeared oand the hilevel was eof sun radiin years wcation thata reason fHowever, tof root sysden decreanot have econtrary, tratively affected by high groundwater level. This can explainhigh decrease in radial increments in very humid and cloudyresult intionsandicaler thul foapping cely towluld lvenvenif Knljanl Ager enc, O.,iablesst, Pr4, 39., 200artmSlovelewskate c112.A., Peth il. Mag, M.ory. G., 20planlandljann, J.Ocentrn, J.Otosyn., BacdelhenerBornbachwarzpp.rstonus glp://wManazdnodoveManazdnoo the model frequent white frosts stimulate radialat low Ledava River levels and high sun radia-ities, whereas increments rapidly fall toward zeroiver levels and lower sun radiation intensities. Thenumber of years with frequent late white frostsst probable reason for the inexplicable behavior ofunder high number of late white frosts.del suggests that white frost is important, but itestroyable for this tree species.peraturetance of temperatures for the radial growth oftrees in this area was found also in the research(1993), whereas Horacek et al. (2003) found thattemperatures, precipitation and soil water sup-portant for the radial increments of a oodplainperature uctuations also importantly affectedth of moist forests in Lithuania (Kairiukstis andne, 1987).radial increments were importantly affected byntal conditions in August, September and October.d quite unusual as the majority of radial growthually takes place until July (Keeland and Sharitz,be et al., 2000; Costa et al., 2003) and it decreaseserwards. However, in black alder trees or seedlingsth can take place even as late as in September or(Eschenbach, 1995; Kaelke andDawson, 2003; Costa. Growth in the second part of the growing seasonounced in treeswith diffuse porouswood andwithgrowth. Their radial growth follows the budburstlder is known to sprout new leaves as late as in(Eschenbach, 1995).nclusionmachine learning algorithmhas proved its advan-sefulness in ecologicalmodeling. The examinationted growthmodel implies somenewndings aboutses of black alder on dynamic changes of environ-ibutes. It was showed that the combination of theer level (and consequently groundwater level) andn of sun radiation have the major inuence on thef radial increments on the selected study site. Aannual radial increments due to the drought stressnly in years with the lowest groundwater levelsghest sun radiation intensities. The groundwaterspecially important in years with high durationation, whereas sun radiation was more importantith high groundwater levels. There was no indi-a decrease in groundwater table alone could beor tree decline on this site in the present time.rees must undergo several important adaptationstem to survive in changed conditions. In a sud-se of groundwater these trees most probably wouldnough time for appropriate adaptations. On theees adapted to lower groundwater levels are neg-years.Theter resoscillastandsphologWe infstressfTheunfoldnot likAcknWe wothe Slothe Sloment oin Ljubmentar e f eAntonivarfore33/3Brus, RDep(inChmieclim101Costa,growEcoCuliberhistCater, MandlowLjubDawsoconDawsophoDilly, OMidandtheEschenSch197Feathe(Aln(httForest(GogozForest(Goults indicate that high oscillations of groundwa-increased oscillations of radial increments. Highin water conditions represent a stress to thesethey demand high-energy investments for mor-and physiological modications of the root high uctuations of groundwater level are mostr these stands.earance of late white frost at the time of leafan importantly affect radial increments, but it iso cause a decline in black alder trees.edgementsike to express our gratefulness to the personnel ofia Forest Service, Regional Unit Murska Sobota, toan Forestry Institute in Ljubljana and to theDepart-owledge Technologies at the Jozef Stefan Institutea. Moreover, we would like to thank the Environ-ncy of Slovenia for providing meteorological data.c e sMarki, A., Hatic, D., 1998/99. Modelling of climaticas a part of dendroecological study in the Repasekodravlje, Croatia. Hrvatski meteoroloski casopis51.5. Dendrology for Foresters. Biotechnical Faculty,ent for Renewable Forest Resources, Ljubljana, p. 408nian).i, F.M., Rotzer, T., 2001. Response of tree phenology tohange across Europe. Agric. Forest Meteorol. 108,reira, H., Oliveira, A., 2003. Variability of radialn cork oak adult trees under cork production. Forestnage 175 (13), 239246., 1989. Forests of Prekmurje in closer and longerozdarski vestnik 47 (5), 218223 (in Slovenian).02. Effect of light and groundwater table on naturalted seedlings of pedunculate oak (Quercus robur L.) inparts of Slovenia. Forestry Institute of Slovenia,a, p. 115 (in Slovenian).., Funk, D.T., 1981. Seasonal change in foliar nitrogenation of Alnus glutinosa. Forest Sci. 27 (2), 239243.., Godon, J.C., 1979. Nitrogen xation in relation tothesis in Alnus glutinosa. Bot. Gaz. 140, S70S75.h, H.J., Buscot, F., Eschenbach, C., Kutsch, W.L.,off, U., Pritsch, K., Munch, J.C., 2000. Characteristicsgetic strategies of the rhizosphere in ecosystems ofhoved Lake district. Appl. Soil Ecol. 15, 201210., C., 1995. Zur Physiologie und Ekologie dererle Alnus glutinosa (L.). Dissertation, Universitat Kiel,e, A.W., 2003. Species prole: Common or black alderutinosa). Caledonia Wild! Newsletters, spring Plan for Forestry Unit Dolinskogospodarski nacrt GE Dolinsko), 19591968. Zavod zaSlovenije, Slovenia Forest Service (in Slovenian).gement Plans for Forestry Unit Dolinskogospodarski nacrti GE Dolinsko), 19711980;ecolog ical modell ing 2 1 5 ( 2 0 0 8 ) 180189 18919811990; 19922001; 20022011. Zavod za gozdove Slovenije,Slovenia Forest Service (in Slovenian).Gordon, J.C., Wheeler, C.T., 1978. Whole plant studies inphotosynthesis and acetylene reduction in Alnus glutinosa.New Phytol. 80 (1), 179186.Grosse, W., Schroder, P., 1986. Panzenleben unter anaerobenUmweltbedingungen, die physikalischen Grundlagen undanatomischen Voraussetzungen. Berichte der DeutschenBotanischen Gesellschaft 99, 367381.Hibbs, D.E., Emmingham, W.H., Bondi, M.C., 1989. Thinning redalder: effects of method and spacing. Forest Sci. 35 (1), 1629.Horacek, P., Slezingerova, J., Gandelova, L., 2003. Analysis ofcambial activity and formation of wood in Quercus roburunder conditions of a oodplain forest. J. Forest Sci. 49 (9),412418.Iremonger, S.F., Kelly, D.L., 1988. The response of four Irishwetland tree species to raised soil water levels. New Phytol.109 (4), 491497.Kaelke, C.M., Dawson, J.O., 2003. Seasonal ooding regimesinuencenitrogen(1), 1671Kairiukstis,moist foecologicaGeogr. 14Kalan, J., 198ResearchManagemSloveniaKeeland, B.Ductuatiswamp.Kosir, Z., 198Crni Logmanagem(in SloveKrstinic, A.,conservathe Fourand the(Turok, J.Rome, ItLaganis, J., 2(L.) Gaertof Nova GLebaube, S.,balanceSci. 57, 4Levanic, T.,incremeGaertn.),L.) in Prekmurje. Master Thesis, Biotechnical Faculty,Ljubljana (in Slovenian, with English abstract).Levanic, T., Kotar, M., 1996. The effect of hydroameliorations onthe forest conditions at Crni Log. In: Drainage andEnvironment, 6th ICID Drainage Workshop, Ljubljana,Slovenia, April 2224.McVean, D.N., 1953. Biological ora of the British Isles: Alnusglutinosa (L.) Moench. J. Ecol. 41 (2), 447466.McVean, D.N., 1956. Ecology of Alnus glutinosa (L.) Moench. IV.Root system. J. Ecol. 44 (1), 219225.Nemesszeghy, L., 1986. Black Alder in Prekmurje. Pomurskazalozba, p. 88 (in Slovenian).Peckov, A., Todorovski, L., Dzeroski, S., 2006. Proper versus ad hocMDL principle for polynomial regression. In: Proceedings ofthe Ninth International Multiconference Information SocietyIS 2006, Ljubljana, Slovenia, October 914.Pizelle, G., 1984. Seasonal-variations of the sexual reproductivegrowth and nitrogenase activity (C2H2) in mature Alnusglutinosa. Plant Soil 78 (1/2), 181188.l, D., Knor, E.M., Reif, A., 1997. Degradation of alder swampst innbruhandman,er, P.,spor44.ltural2200t Dol, I., 1s. Oaitutevski,ationrnatilin, p, R.Hcienr, C.Ttion495.ead,tran644., M.,orest230, I., 1ertova Risurvival, nitrogen xation, and the partitioning ofand biomass in Alnus incana ssp rugosa. Plant Soil 25477.L.A., Stravinskiene, V.P., 1987. Dendrochronologies forrests of the Lithuanian SSR and their application forl forecasting. Ann. Acad. Sci. Fenn. Ser. A III: Geol.5, 119135.8. Pedological Conditions in the Forests of Crni Logwork. Institute for Forestry and Silvicultureent at Biotechnical Faculty, Ljubljana, p. 21 (inn).., Sharitz, R.R., 1997. The effects of water-levelons on weekly tree growth in a southeastern USAAm. J. Bot. 84 (1), 131139.7. Effects of the Radmozanci Dam on the Forests ofResearch work. Institute for forestry and woodent at Biotechnical Faulty, Elaborate 244, Ljubljananian).Gracan J., Kajba D., 2002. Alder genetic resourcestion strategy. Noble Hardwoods Network, Report ofth meeting, September 46, 1999; Gmunden, SustriaFifth meeting, May 1719, 2001, Blessington, Ireland, Erikson, G., Russel, K., Borelli, K., compilers). IPGRI,aly, pp. 4449.007. Emergy analysis of black alder (Alnus glutinosan.) oodplain forest growth. Dissertation. Universityorica, 175 p.Le Goff, N., Ottorini, J.M., Granier, A., 2000. Carbonand tree growth in a Fagus sylvatica stand. Ann. Forest961.1993. Effects of hidromelioration on growth andnt characteristics of black alder (Alnus glutinosa (L.)ash (Fraxinus angustifolia Vahl.) and oak (Quercus roburPretzelforeErleVerGerSchrodtran38Silvicu199UniSmolejplotInstTodoroequInteBerWaringBiosWheelexa487Whitehand633Wraberin f179ZibernaIn: BDrathe upper Rhine Valley (Degradation vonchwaldern in der Oberrheinebene).lungen-Gesellschaft-fur-Okologie 27, 435440 (inwith English abstract).1989. Characterization of a thermo-osmotic gast mechanism in Alnus glutinosa (L.) Gaertn. Trees 3,chronicles, 19711980, 19721981, 19821991,1. Silvicultural Area 13, Murska Sobota, Silviculturalinsko, Lendava.995. Hydrological conditions on permanent researchk decline in Slovenia, Final Report, Ljubljana, Forestryof Slovenia, 37 pp.L., Ljubic, P., Dzeroski, S., 2004. Inducing polynomials for regression. In: Proceedings of the 15thonal Conference on Machine Learning, Springer,p. 441452.., 1987. Characteristics of trees predisposed to die.ce 37, 569574.., 1971. The causation of diurnal changes in nitrogenin the nodules of Alnus glutinosa. New Phytol. 70 (3),D., 1998. Regulation of stomatal conductancespiration in forest canopies. Tree Physiol. 18,1951. The review on forest vegetation and problemsmanagement in Prekmurje. Geografski vestnik 23,(in Slovenian).992. Climatic description of Northeastern Slovenia.ncelj, D. (Ed.), International Conference about thever, pp. 5562 (in Slovenian).Modeling radial growth increment of black alder (Alnus glutionsa (L.) Gaertn.) treeIntroductionMaterials and methodsStudy siteSite descriptionClimate and soil conditionsHistoryMethodsDatasetData analysisModelingResultsDiscussionAgeThe Ledava River level and duration of sun radiationThinningWhite frostTemperatureConclusionAcknowledgementsReferences


View more >