e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208

avai lab le at www.sc iencedi rec t .com

journa l homepage: www.e lsev ier .com/ locate /eco lmodel

Modelling the damage status of silver fir trees (Abies albaMill.) on the basis of geomorphological, climatic andstand factors

Mario Bozic a,∗, Oleg Antonic b, Renata Pernara, Sven D. Jelaskac, Josip Krizanc,Juro Cavlovic a, Vladimir Kusanc

a Faculty of Forestry, University of Zagreb, Svetosimunska 25, 10000 Zagreb, Croatiab Ru –der Boskovic Institute, Bijenicka 54, 10000 Zagreb, Croatiac Oikon Ltd. Institute for Applied Ecology, Vlade Prekrata 20, 10020 Zagreb, Croatia

a r t i c l e i n f o a b s t r a c t

Article history:

Available online 28 November 2005

Keywords:

Digital elevation model

Fir health status

General linear modelling

Macroclimate

Raster-GIS

The paper investigates the possibility of assessing damage (health status) of silver fir (Abies

alba Mill.) using multivariate regression models in the function of geomorphological, cli-

matic and stand factors. Research was carried out in beech-fir forests (Abieti-Fagetum), on

limestone-dolomite substrate, in the Dinaric part of the silver fir range in the Republic of

Croatia. A general linear modelling method was used, where square terms and interaction

terms (multiplication products) of original variables were treated as independent linear pre-

dictors. Twenty-seven separate models (all of them with significant part of explained fir

damage variability) were developed regarding the different subsets of input data, different

subsets of independent variables and different model design. Some of these models could be

preliminary used for spatial predicting and mapping of fir damage in a frame of raster-GIS,

for entire area of Dinaric part of Abieti-Fagetum in Croatia.

© 2005 Elsevier B.V. All rights reserved.

1. Introduction

Forests are permanently influenced by numerous biotic andabiotic factors. Some of these lead to the weakening of lifeforce (vitality) and dieback of forest trees. Forest conditionsdepend on soil, tree age, climate, pests and diseases and othernatural stressors (Aamlid et al., 2000).

Beech-fir forests (Abieti-Fagetum) are the most widely dis-tributed altimontane forests on Croatian karst, taking upan area of over 300 000 ha (Trinajstic et al., 1992). Theseare uneven-aged forests (stands containing trees of differentheights, diameters and ages over a unit area) managed on aselection basis. Accounting for 10% in the total growing stock,fir is one of the most important tree species (and the most

∗ Corresponding author. Tel.: +385 1 2352498; fax: +385 1 2352514.E-mail address: bozic@sumfak.hr (M. Bozic).

important conifer species) in Croatia (Mestrovic, 2001). Forsome time now silver fir has been faced with growing damageand decline. The direct consequence of increased tree damage(decreased vitality) is reflected in a reduced increment (Bert,1993; Bezak et al., 1991; Kalafadzic and Kusan, 1989; Klepac,1972; Pernar et al., 2000; Prpic and Seletkovic, 1992; Prpic etal., 1994; Schopfer and Hradetzky, 1986). An increase in thedamage degree reduces the variability of diameter increment,which means that damaged trees show a weakened reactionto environmental effects (Kalafadzic and Kusan, 1989).

Tomasevski (1958) puts the cause of intensive dieback offir in Croatia immediately after the Second World War downto chaotic felling, which disturbed the biocoenological bal-ance. Physiologically debilitated trees are prone to attacks by

0304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolmodel.2005.10.021

e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208 203

secondary pests: bark beetles (Tomasevski, 1958) andArgyresthia fundella (Androic and Klepac, 1969; Klepac, 1972).Prpic (1987) links their gradation to warming up and decreasedhumidity in the sites of fir forests. In recent times, a numberof authors have connected increased dieback of forest trees(especially fir) in Croatia with the impact of harmful pollu-tants (Glavac et al., 1985; Kauzlaric, 1988; Plese-Lukeza, 1997;Prpic, 1987), similar to other authors in different countries(Aamlid et al., 2000; Innes and Cook, 1989; Lagana et al., 2000;Nellemann et al., 2003; Oszlany, 1997). The highest degreeof damage is seen in trees on massifs exposed to prevailingwinds (Bert, 1993), which probably bring pollutants (Antonicand Legovic, 1999). The real causes of fir dieback in Croatia arestill unknown, and so are the causes of decline of some othertree species (see, e.g. Akashi and Mueller-Dombois, 1995).

Health condition of stands is also affected by geomorpho-logical factors (terrain aspect, slope and altitude). In his studyof fir tree dieback, Tomasevski (1958) concluded that diebackwas more intensive on the warmer terrain orientations thanon the colder ones. Safar (1951) and Prpic and Seletkovic (1992)reached similar conclusions relating to more intensive dam-age in warm positions. A higher terrain slope and altitude alsoincrease damage to fir trees (Kalafadzic et al., 1992; Prpic et al.,1991; Seletkovic and Potocic, 2001; Safar, 1951). In none of theabove research into dieback of fir trees in Croatia were geomor-phological variables used as predictors for the assessment ofthe damage degree.

a(diatddtoiwl2

This paper investigates the possibility of assessing andmaking spatial predictions of the damage (health) status ofsilver fir in Croatian karst beech-fir forests as the function ofgeomorphological, climatic and stand factors (especially in thelight of recent experiences with the application of raster-GISmodelling in ecological research of karst areas (Antonic andLegovic, 1999; Antonic et al., 2000, 2001a, 2003)).

2. Materials and methods

2.1. Study area

Research was carried out within the area of beech-fir forests(Fig. 1a) on limestone-dolomite substrate in the Dinaric partof the silver fir range in the Republic of Croatia (followingTrinajstic et al., 1992), as most frequent, economically mostimportant and most endangered (regarding the fir healthstatus) forest type with fir as dominant (or co-dominant)species.

2.2. Forest damage data

Data used as dependent variables for model development werecollected with field sampling in 151 plots (Fig. 1b). One tofour trees of varying breast diameters (10–30, 30–50, 50–70 and>70 cm) closest to the plot centre were sampled in each plot.

F este rest

Olthof et al. (2004) and Zierl (2004) investigate forest dam-ge on the basis of climatic indicators. Dobbertin and Brang2001) prove the correlation between tree mortality and treeamage of the previous year and the social position of a tree

n the stand, whereby the impact of competition is regarded asstress factor. According to Heski (1990), the thickest and thin

rees (in a selection forest) are less resistant to the impact ofefoliation-related factors than medium-thick trees. Greateramage to thin trees is probably the consequence of competi-ion, whereas greater damage to thick trees is the consequencef exposure to air currents and air pollutants. Tree age (which

n a fir, as a sciophylic species is not necessarily correlatedith tree thickness) is an important factor that affects defo-

iation (Klap et al., 2000; Mayer, 1999; Pouttu and Dobbertin,000; Seletkovic and Potocic, 2001; Safar, 1965).

ig. 1 – (a) Dinaric part of the silver fir range (grey area) on limt al., 1992); (b) position of groups of sample plots and ICP fo

The plots were organized in transects, traced to cover existinglocal variability of geomorphological variables (altitude, ter-rain aspect and slope). A total of 531 trees of silver fir (Abiesalba Mill.) were sampled.

To perform the damage assessment, tree crowns were com-pared with the existing photo-interpretation key (Anon., 1989;Kusan et al., 1991), on the damage scale of 5%.

In addition, data collected independently of this researchwithin the International Cooperative Programme on Assess-ment and Monitoring of Air Pollution Effects in Forests (ICPForest) were compared with the results obtained by spatialpredictions using developed models. Data from the points of4 km × 4 km and 16 km × 16 km networks were used, in whichthere were ≥6 (25%) of fir trees. The mean damage of fir treeswas calculated.

one-dolomite substrate in the Republic of Croatia (Trinajsticplots.

204 e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208

2.3. Independent variables

Independent variables used as a basis for the performed mod-elling of spatial damage distribution were divided into threegroups: (1) climatic variables (monthly mean air temperature,monthly precipitation, monthly mean global solar irradiationon horizontal surface at ground level and monthly potentialevapotranspiration on horizontal surface) as macroclimaticpredictors; (2) geomorphological variables, all of them derivedfrom a digital elevation model (DEM) in spatial resolutionof 100 m × 100 m (altitude, terrain slope, terrain aspect, flowaccumulation, sinkhole depths, terrain curvature indicators,terrain exposure to the horizontal wind flux), as topoclimaticpredictors (Antonic, 1996; Antonic et al., 2001a) or as predictorsof pollutant-related damage (Antonic and Legovic, 1999); (3)stand variables (diameter breast height (DBH), tree basal areain relation to plot basal area (g/G) and tree age). Apart fromthe three mentioned groups of variables, latitude and longi-tude (transformed into the local co-ordinate system) were alsoused as potential additional estimators of macroclimatic gra-dients (uncovered by above mentioned climatic variables; e.g.distance from major cyclonal path, see Robinson, 1984).

The first-two climatic variables (monthly mean air tem-perature, monthly precipitation) were taken directly from theweather station chronicles. The last two were modelled foreach weather station as a function of relevant climatic vari-ables observed at the respective station, based on Nikolov and

Tabl

e1

–S

um

mar

yre

sult

sfo

ral

lmod

els

Use

dva

riab

les

Pred

icto

rsIn

clu

ded

tree

sD

EM-b

ased

vari

able

sD

EM-b

ased

and

stan

dva

riab

les

wit

hou

tag

eD

EM-b

ased

and

stan

dva

riab

les

Nr.

R2

N2P

+1

F%

Nr.

R2

N2P

+1

F%

Nr.

R2

N2P

+1

F%

All

DEM

-bas

edva

riab

les

Lin

ear,

squ

ares

,in

tera

ctio

ns

All

10.

422

531

185

3.47

659

.82

20.

520

531

235

3.83

031

.62

30.

542

531

255

3.75

531

.35

≥40

cm4

0.69

826

118

34.

285

39.3

05

0.74

726

120

34.

660

29.8

66

0.82

026

123

35.

672

34.9

1<

40cm

70.

468

270

123

2.99

571

.10

80.

669

270

229

2.74

323

.76

90.

634

270

201

2.93

243

.89

All

DEM

-bas

edva

riab

les

Lin

ear,

inte

ract

ion

sA

ll10

0.44

153

120

13.

386

64.3

911

0.49

653

119

94.

276

66.4

512

0.52

553

121

54.

369

60.5

7≥4

0cm

130.

679

261

183

3.92

657

.35

140.

702

261

167

5.10

363

.30

150.

800

261

231

5.04

223

.36

<40

cm16

0.56

127

016

72.

865

55.7

017

0.69

527

023

92.

878

27.6

118

0.77

327

028

53.

047

55.7

0

Red

uce

dse

tof

DEM

-bas

edva

riab

lesa

Lin

ear,

squ

ares

,in

tera

ctio

ns

All

190.

301

531

994.

230

91.9

920

0.35

153

197

5.43

193

.19

210.

403

531

141

4.44

485

.10

≥40

cm22

0.46

526

193

4.03

689

.60

230.

573

261

117

4.66

686

.25

240.

557

261

107

4.91

390

.92

<40

cm25

0.42

927

011

52.

794

72.6

326

0.40

427

093

3.28

972

.63

270.

425

270

973.

402

84.3

5

See

Sect

ion

2.3

for

des

crip

tion

ofin

dep

end

ent

vari

able

s.N

r.,m

odel

nu

mbe

r(b

old

nu

mbe

rsre

pre

sen

tm

odel

sse

lect

edfo

rsp

atia

lpre

dic

tion

ofsi

lver

fir

dam

age

stat

us

atth

een

tire

rese

arch

area

);R

2,

det

erm

inat

ion

coef

fici

ent;

N,n

um

ber

oftr

ees;

P,n

um

ber

ofp

aram

eter

sin

the

mod

el;%

,por

tion

ofva

lues

pre

dic

ted

wit

hin

the

ran

geof

inp

ut

dat

aon

fir

dam

age

(0–1

00%

)at

the

enti

reD

inar

icp

art

ofA

biet

i-Fa

getu

m;p

roba

bili

tyof

F-st

atis

tics

was

less

than

0.00

0in

allm

odel

s.a

Lim

ited

nu

mbe

rof

DEM

-bas

edva

riab

les

pre

sum

edas

the

mos

tim

por

tan

tes

tim

ator

sof

tree

dam

age:

mon

thly

mea

nai

rte

mp

erat

ure

,mon

thly

pre

cip

itat

ion

,mon

thly

mea

ngl

obal

sola

rir

rad

iati

onon

the

hor

izon

tal

surf

ace

atgr

oun

dle

vel

and

mon

thly

pot

enti

alev

apot

ran

spir

atio

non

the

hor

izon

tal

surf

ace,

terr

ain

slop

e,al

titu

de,

terr

ain

asp

ect,

terr

ain

exp

osu

reto

the

hor

izon

tal

win

dfl

ux

and

lati

tud

ean

dlo

ngi

tud

eas

add

itio

nal

ind

epen

den

tva

riab

les.

Zeller (1992) for global solar irradiation and Priestly and Taylor(1972) and Bonan (1989) for potential evapotranspiration. Spa-tial distributions of climatic variables were averaged for theperiod 1956–1995 in spatial resolution of 300 m × 300 m, usinginterpolation models presented in Antonic et al. (2001b). Thesemodels were developed with neural networks (NN), based ondata from 127 weather stations and elevation data from DEM.

Since the terrain aspect is a circular variable, it is con-sequently transformed into northness and eastness by thecosine and sine transformation, respectively (Guisan et al.,1998, 1999). Northness was used as a general estimator of localthermics, while eastness (together with terrain exposure tothe horizontal wind flux from the west (Antonic and Legovic,1999; Matocec et al., 2000). Altitude, latitude and longitudewere adopted in this research as an indirect estimator of forestcontamination with aeropollutants. Sinkhole depths (Antonicet al., 2001a) were included in the analysis as estimators oflocal air temperature and humidity, due to the confluence ofthe heavier air (colder and/or more humid) in the sink (so-called temperature inversion). Flow accumulation, calculatedas a logarithm of the number of pixels in DEM that gravitateto the given pixel (see, e.g. Beven and Kirkby, 1979; Burt andButcher, 1985), was included as a potential estimator of soilmoisture. Terrain curvature (the rate of change of the surfaceslope in a given direction) was included in research with twovariables: (1) in the direction of maximal terrain slope (profilecurvature; affects the acceleration and deceleration of flowand, therefore, influences erosion and deposition) and (2) inthe direction perpendicular to the direction of the maximalterrain slope (planform curvature; influences the convergenceand divergence of flow).

DBH was used as the estimator of social position of a par-ticular tree in the forest, while g/G was used as the estimator

e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208 205

of competitors’ pressure. Tree age was also used as an inde-pendent variable due to the fact that it is not necessarily corre-lated with DBH in uneven-aged forests, especially those withsciophytic tree species like a silver fir. Stand variables werecollected in sample plots. Plot basal area (G) was calculatedas a sum of basal areas of individual trees (excluding thosewith DBH smaller than 10 cm) in a circular plot with a radiusof 12.5 m. Tree age (above breast height) was calculated fromlong tree cores taken at breast height.

2.4. Models

A general linear modelling method was applied where squareterms and interaction terms (multiplication products) of origi-nal independent variables were treated as independent linearpredictors (Ott, 1993) of fir damage estimation. The ‘BackwardStepwise’ method (Ott, 1993) was used for model optimisa-tion (selection of a subset of linear predictors entering thefinal model). All statistical processing retained the significancelevel of p = 0.05.

During modelling, care was taken that the total number ofestimators in the model (including all original variables, theirsquares and interactions) satisfies the condition: 2P + 1 < N(Ott, 1993), where P is the number of parameters (estimators)in the model, and N is the sample size (number of trees).

Twenty-seven separate models (see also Table 1) weredeveloped regarding the different subsets of input data (alltbiiad

3

Mma

variability of damage status of silver fir at the research area.Both models developed from separate data sets with regard todiameter of 40 cm (presumed a boundary between dominanttrees and suppressed or young trees) explained a significantlylarger part of total variability (in all combinations of othermentioned model characteristics) in relation to the respec-tive models developed for all trees together. Consequently,models developed for all trees were not examined any fur-ther, presuming that processes leading to forest dieback aresignificantly influenced by the social position of a tree.

All the remaining models were preliminarily used for theconstruction of hypothetical spatial distributions of fir dam-age for the entire fir area in Croatia. The criteria for the finalselection of models potentially applicable to spatial predic-tion of fir damage was combination of: (1) total variabilityexplained by the model (R2) and (2) a portion of predicted val-ues within the range of input data on fir damage (0–100%).

The selected models were applied to entire Dinaric partof Abieti-Fagetum area (Trinajstic et al., 1992), aiming at con-structing a hypothetical spatial distribution of silver fir healthstatus. This was done within the frame of raster geographicsystem (raster-GIS) using all geomorphological and climaticestimators in the spatial resolution of 200 m × 200 m.

For raster pixels without actual data for stand variables,mean values were used in those models that use these vari-ables. Fig. 2 shows spatial distribution of estimated silverfir health status for two developed models (with and with-

F treesD 0–25>

rees, trees with diameter over 40 cm and trees with diameterelow 40 cm), different subsets of independent variables (lim-

ted number of DEM-based variables presumed as the mostmportant estimators of tree damage, all DEM-based variables,ll variables without age, all variables) and different modelesign (with and without square terms).

. Results and discussion

odelling results are summarised in Table 1. Parameters’ esti-ations for particular models were omitted and they are

vailable on request. All models explain significant part of

ig. 2 – Potential spatial distribution of mean fir damage foramage values are sorted in six classes: <10%, dark green; 180%, black.

out stand factors included as an estimators). Without moredata from the field it is hard to tell whether obvious differ-ences between two models presents better prediction withstand structures data included in the model (that higherR2 suggest—Table 1) or drawback of using mean values forpixels without ground truth data. Answering that questionwill be one of the directions in our future research on thetopic.

The interpretation of unexplained variability and a portionof predicted values in the expected range in all developedmodels (Table 1) could be addressed to: (1) the static natureof the model (the registered damage did not occur at once; cli-matic aberrations from the average were not included in the

with diameters ≥40 cm for models 22 (left), and 24 (right).%, green, 25–40%, yellow; 40–60%, orange; 60–80%, red;

206 e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208

Table 2 – Correlation between predicted damage frommodels 22–27 and ICP Forest mean damage

Model

22 23 24 25 26 27

R −0.2726 −0.2006 −0.1753 0.5096 0.4269 0.5360P p = 0.131 p = 0.271 p = 0.337 p = 0.003 p = 0.015 p = 0.002

model (see, e.g. Zierl, 2004); (2) mechanically incurred damage(damage due to felling, tree logging, wind and snow impacts);(3) damage resulting from primary and secondary pest out-breaks (insects, see e.g. Androic and Klepac, 1969; Klepac,1972 or fungi; (4) regularly applied management procedures(removing more severely damaged trees (Prpic et al., 1994; seealso Oszlany, 1997) in some localities, which may decrease themean tree damage in a plot (see, e.g. Pernar and Kusan, 2001;Seletkovic and Tikvic, 1996); (5) the impacts of local pollutantsources, for example, along the roads (see, e.g. Ekstrand, 1994);(6) level to which sampled plots represents whole modelledarea as observed in Kohl and Gertner (1997); (7) insufficientmodel complexity.

The six finally selected models (all with a reduced set ofDEM-based variables and square terms; models 22–27) weretested on the independent data set collected within the Inter-national Cooperative Programme on Assessment and Moni-toring of Air Pollution Effects (ICP) on Forests (Anon., 1989). Ithas to be emphasized that ICP data cannot be interpreted inthis research as fully reliable ground truth, due to the knownproblem of visual interpretation (especially in a case of largernumber of field observers, which took a place here), high-lighted by a number of authors (Bednarova, 2001; Dobbertinand Brang, 2001; Wilksch et al., 1998; Wohlfahrt et al., 1998).Beside those limitations, ICP data remains only existing fielddata, hence, acceptable for model validation.

A significant correlation between model results was

modelling techniques such as Neural Networks, Support Vec-tor Machines and Naıve Bayesian Classifiers.

Acknowledgements

This work was supported by Hrvatske Sume Ltd., OIKON Ltd.Institute for Applied Ecology and Ministry of Science, Edu-cation and Sport of Republic of Croatia. Three anonymousreviewers helped us with their comments and suggestions toclarify the text and make the complete paper better.

r e f e r e n c e s

Aamlid, D., Torseth, K., Venn, K., Stuanes, A.O., Solberg, S.,Hylen, G., Christophersen, N., Framstad, E., 2000. Changes offorest health in Norwegian boreal forests during 15 years.For. Ecol. Manage. 127, 103–118.

Akashi, M., Mueller-Dombois, D., 1995. A landscape perspectiveof the Hawaiian rain forests dieback. J. Veg. Sci. 6, 449–464.

Androic, M., Klepac, D., 1969. The problem of silver fir diebackin the areas of Gorski Kotar, Lika and Slovenia (in Croatianwith an English abstract). Sumarski List 1–2, 1–12.

Anon., 1989. Manual on Methodologies and Criteria forHarmonized Sampling, Assessment, Monitoring andAnalysis of the Effects of Air Pollution on Forest. UnitedNations Environ. Program (UNEP) and United Nations

obtained only for trees thinner than 40 cm (Table 2). Absenceof significant correlations for trees thicker than 40 cm can beexplained by the impact of forest management that favoursthe removal of damaged mature trees, which is additional rea-son for partial unreliability of ICP data, as well as fir damagedata collected in this research. This could be important inputfor future collecting of data in the dependence of forest man-agement practice.

4. Conclusions

The obtained results could be preliminarily used for spatialprediction and mapping of fir damage within raster-GIS for theentire area of Abieti-Fagetum in Croatian karst. Future researchshould focus on: (1) completing a larger field sample so thatmore plastic prediction models can be provided (e.g. developedby neural networks); (2) integrating spatial forest health pre-diction models with spatial models of aeropollutant immis-sions (Antonic and Legovic, 1999) and other relevant factors,e.g. impact of pests and mechanically incurred damage; (3)including the temporal aspect of forest health; (4) using ofmore reliable ground truth data; (5) using of more advanced

Economic Commission for Europe (UN-ECE). Fed. Res. Centrefor For. and For. Products, Hamburg.

Antonic, O., 1996. Application of spatial modeling in the karstbioclimatology. Cro. Met. J. 31, 95–102.

Antonic, O., Legovic, T., 1999. Estimating the direction of anunknown air pollution source using a digital elevationmodel and a sample of deposition. Ecol. Model. 124, 85–95.

Antonic, O., Bukovec, D., Krizan, J., Marki, A., Hatic, D., 2000.Spatial distribution of major forest types in Croatia as afunction of macroclimate. Nat. Croat. 9, 1–13.

Antonic, O., Hatic, D., Pernar, R., 2001a. DEM-based depth insink as an environmental estimator. Ecol. Model. 138,247–254.

Antonic, O., Krizan, J., Marki, A., Bukovec, D., 2001b.Spatio-temporal interpolation of climatic variables over alarge region of complex terrain using neural networks. Ecol.Model. 138, 255–263.

Antonic, O., Pernar, N., Jelaska, S.D., 2003. Spatial distributionof main forest soil groups in Croatia as a function of basicpedogenetic factors. Ecol. Model. 170, 363–371.

Bednarova, E., 2001. Reaction of assimilatory tissues of Norwayspruce (Picea abies [L.] Karst.) and silver fir (Abies alba Mill.)to immission stress in regions exposed to a long-term airpollution load. Ekol. Bratislava 20, 36–45.

Bert, G.D., 1993. Impact of ecological factors, climatic stressesand pollution on growth and health of silver fir (Abies albaMill.) in the jura mountains—an ecological anddendrochronological study. Acta Oecol. Int. J. Ecol. 14,229–246.

Beven, K.J., Kirkby, M.J., 1979. A physically based variablecontributing area model of basin hydrology. Hidrol. Sci. Bull.24, 43–69.

Bezak, K., Krejci, V., Vrbek, B., 1991. Fir dieback accompaniedby changes in the vitality and increment of beech and firforests in the period 1969–1990. Radovi 26, 115–128.

Bonan, G.A., 1989. Computer model of solar radiation, soilmoisture and soil thermal regimes in boreal forests. Ecol.Model. 45, 275–306.

e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208 207

Burt, T.P., Butcher, D.P., 1985. Topographic controls of soilmoisture distributions. J. Soil Sci. 36, 469–486.

Dobbertin, M., Brang, P., 2001. Crown defoliation improves treemortality models. For. Ecol. Manage. 141, 273–286.

Ekstrand, S., 1994. Close range forest defoliation effects oftraffic emissions assessed with aerial photography. Sci. TotalEnviron. 147, 149–155.

Glavac, V., Koenies, H., Prpic, B., 1985. On the entry of aerealpollutants into beech and fir forests on the dinaric range ofsouth-western Yugoslavia (in Croatian with Englishabstract). Sumarski List 9–10, 429–447.

Guisan, A., Theurillat, J.-P., Kienast, F., 1998. Predicting thepotential distribution of plant species in an alpineenvironment. J. Veg. Sci. 9, 65–74.

Guisan, A., Weiss, S.B., Weiss, A.D., 1999. GLM versus CCAspatial modeling of plant species distribution. Plant Ecol.143, 107–122.

Heski, T., 1990. Dieback rate of fir Trees in Gorski Kotar as aconsequence of trade (in Croatian with English abstract).Sumarski List 6–8, 289–294.

Innes, J.L., Cook, E.R., 1989. Tree-ring analysis as an aid toevaluating the effects of pollution on tree growth. Can. J.For. Res. 19, 1174–1189.

Kalafadzic, Z., Kusan, V., 1989. Decrease in fir (Abies alba L.)increment as a result of forest decline in Gorski Kotar (inCroatian with English abstract). Sumarski List 9–10, 415–422.

Kalafadzic, Z., Kusan, V., Horvatic, Z., Fintic, R., 1992. DerGesundheitszustand der Tanne (Abies alba Mill.) inSudwesten Kroatiens aufgrund derFarb-Infrarot-Luftbildinterpretation (in German). In: Prpic, B.,

In: Karnosky, D.F., Percy, K.E., Chappelka, A.H., Simpson, C.,Pikkarainen, J., (Eds.), Air Pollution, Global Change, Forestsin the New Millennium. pp. 277–288.

Nikolov, N.T., Zeller, K.F., 1992. A solar radiation algorithm forecosystem dynamic models. Ecol. Model. 61, 149–168.

Olthof, I., King, D.J., Lautenschlager, R.A., 2004. Mappingdecidous forest ice storm damage using Landsat andenviromental data. Remote Sens. Environ. 89, 484–496.

Oszlany, J., 1997. Forest health and environmental pollution inSlovakia. Environ. Pollut. 98, 389–392.

Ott, R.L., 1993. An Introduction to Statistical Methods and DataAnalysis. Duxbury Press, Belmont, p. 1051.

Pernar, R., Kusan, V., Galic, Z., 2000. Relationship betweendiameter increment and damage status of penduculate oak(Quercus robur L.). Glas. sum. pokuse 37, 241–250.

Pernar, R., Kusan, V., 2001. Monitoring beech and fir forestcondition using color infrared aerial photographs (in Croatianwith English abstract). In: Matic, S., Krpan, A.P.B., Gracan,J. (Eds.), Znanost u potrajnom gospodarenju hrvatskimsumama. Faculty of Forestry, University of Zagreb andForest Research Institute Jastrebarsko, Zagreb, pp. 457–463.

Plese-Lukeza, I., 1997. The ecological issues of Gorski Kotar (inCroatian with English abstract). Sumarski List 7–8, 415–424.

Pouttu, A., Dobbertin, M., 2000. Needle-retention and densitypatterns in Pinus sylvestris in the Rhone Valley ofSwitzerland: comparing results of the needle-trace methodwith visual defoliation assessments. Can. J. For. Res. 30,1973–1982.

Priestly, C.H.B., Taylor, R.J., 1972. On the assessment of surfaceheat flux and evaporation using large-scale parameters.

Seletkovic, Z. (Eds.), Proceedings of the Sixth IUFROTannensymposium. Forstliche Fakultat Universitat Zagreb,pp. 219–231.

Kauzlaric, K., 1988. The effect of damaging pollutants ondieback of forests in Gorski Kotar (in Croatian with Englishabstract). Sumarski List 5–6, 229–244.

Klap, J.M., Oude Voshaar, J.H., De Vires, W., Erisman, J.W., 2000.Effects of environmental stress on forest crown condition inEurope. Part IV. Statistical analysis of relationships. WaterAir Soil Pollut. 119, 387–420.

Klepac, D., 1972. Investigations of the effect of defoliators onthe increment of Fir forests (in Croatian with Englishabstract). Sumarski List 1–2, 40–62.

Kohl, M., Gertner, G., 1997. Geostatistics in evaluating forestdamage surveys: considerations on methods for describingspatial distributions. For. Ecol. Manage. 95, 131–140.

Kusan, V., Kalafadzic, Z., Hrasovec, B., Diminic, D., Zdjelar, M.,1991. Appearance of damaged firs, spruces and beeches.Mala ekoloska biblioteka, Knjiga 5. Hrvatsko ekoloskodrustvo Faculty of Forestry, University of Zagreb and ForestResearch Institute Jastrebarsko, Zagreb, p. 92.

Lagana, A., Salerni, E., Barluzzi, C., Perini, C., De Dominicis, V.,2000. Mycocoenology in Abies alba Miller woods ofCentral-Southern Tuscany (Italy). Acta Soc. Bot. Pol. 69,293–298.

Matocec, N., Antonic, O., Mrvos, D., Piltaver, A., Hatic, D.,Bukovec, D., 2000. Fir forest health estimation based onmycobioindication: a case study for Kriz stream catchment,Gorski kotar, Croatia. Nat. Croat. 9, 15–33.

Mayer, F.J., 1999. Beziehungen zwischen der Belaubungsdichteder Waldbaume und Standortsparametern—Auswertung derbayerischen Waldzustandsinventuren. Forstliche Forsch.Munchen 177, 194.

Mestrovic, S., 2001. Managing forest of silver fir (in Croatianwith English abstract). In: Prpic, B. (Ed.), Obicna jela (Abiesalba Mill. ) u Hrvatskoj. Akademija sumarskih znanosti,Zagreb, pp. 529–560.

Nellemann, C., Thomsen, M.G., Soderberg, U., Hansen, K., 2003.Forest growth and critical air pollutant loads in Scandinavia.

Mont. Weather Rev. 100, 81–92.Prpic, B., 1987. Decline of forest trees in Croatia with special

emphasis on the load of “Gorski Kotar” area with acid rainand heavy metals (in Croatian with English abstract).Sumarski List 1–2, 53–60.

Prpic, B., Seletkovic, Z., Ivkov, M., 1991. Forest decline inCroatia and its relation to biotical and abiotical factorstoday and in the past (in Croatian with German abstract).Sumarski List 3–5, 107–129.

Prpic, B., Seletkovic, Z., 1992. Radial increment of the fir in theFaculty Forest of Zalesina as to climatic excesses and inputof pollutants (in German with English abstract). In: Prpic, B.,Seletkovic, Z. (Eds.), Proceedings of the Sixth IUFROTannensymposium. Forstliche Fakultat Universitat Zagreb,pp. 173–182.

Prpic, B., Seletkovic, Z., Ivkov, M., 1994. Crown damage ofprincipal tree species in Croatia in relation to radialincrements (in Croatian with German abstract). SumarskiList 1–2, 3–10.

Robinson, E., 1984. Dispersion and fate of atmosphericpollutants. In: Treshow, M. (Ed.), Air Pollution and Plant Life.John Wiley, Chicester, pp. 15–37.

Safar, J., 1951. Dieback and regeneration of fir in selectionforests of Gorski Kotar. Sumarski List 8–10, 299–303.

Safar, J., 1965. Das problem des tannensterbens und die art derWaldbewirtschaftung auf dem Macelj-Gebirge (in Croatianwith German abstract). Sumarski List 1–2, 1–16.

Schopfer, W., Hradetzky, J., 1986. Zuwachsruckgang inerkrankten Fichten- und Tannenbestanden.Auswertungsmethoden und Ergebnisse. Forstwiss. Centralbl.105, 446–470.

Seletkovic, Z., Tikvic, I., 1996. Damage of different forestecosystem stands in Croatia (in Croatian with Englishabstract). In: Sever, S., (Ed.), Zastita suma i pridobivanjedrva. Faculty of Forestry, University of Zagreb and ForestResearch Institute Jastrebarsko, Zagreb, pp. 411–422.

Seletkovic, I., Potocic, N., 2001. Comparative survey of crowndamage degree on plots of bioindication network in the

208 e c o l o g i c a l m o d e l l i n g 1 9 4 ( 2 0 0 6 ) 202–208

Republic in Croatia (in Croatian with English abstract). In:Matic, S., Krpan, A.P.B., Gracan, J., (Eds.), Znanost upotrajnom gospodarenju hrvatskim sumama. Faculty ofForestry, University of Zagreb and Forest Research InstituteJastrebarsko, Zagreb, pp. 457–463.

Tomasevski, S., 1958. Which diameter classes and whichexpositions contain the largest number of dead fir trees anddown timber? (in Croatian with English abstract). SumarskiList 7–9, 278–281.

Trinajstic, I., Raus, Ð., Vukelic, J., Medvedovic, J., 1992. In: Raus,Ð. (Ed.), Map of Forest Communities in the Republic ofCroatia. Sume u Hrvatskoj, Zagreb.

Wilksch, W., Schmitt, V., Wild, A., 1998. Ethylene-biosynthesisin conifers—investigations on the emission of ethylene andthe content of ACC and MACC in Norway spruce (Picea abies)and silver fir (Abies alba). Chemosphere 36, 883–888.

Wohlfahrt, S., Schmitt, V., Wild, A., 1998. Investigation onphosphoenol pyruvate carboxylase and proline in damagedand undamaged needles of Picea abies and Abies alba.Chemosphere 36, 877–881.

Zierl, B., 2004. A simulation study to analyse the relationsbetween crown condition and drought in Switzerland. For.Ecol. Manage. 188, 25–38.