The Effects of Car Traffic on Breeding Bird Populations in Woodland
Transcription
The Effects of Car Traffic on Breeding Bird Populations in Woodland
0047486 Journalof Applied Ecology 1995, 32, 187-202 The effectsof car trafficon breedingbird populations in woodland.III. Reductionofdensityin relationto the proximity of main roads RIEN REIJNEN*, JOHAN THISSEN*t RUUD FOPPEN*, CAJO TER BRAAK*t and * DLO-Institutefor Forestryand NatureResearch,Departmentof Landscape Ecology, PO Box 23, 6700 AA Wageningen,The Netherlandsand tIDLO-Agricultural MathematicsGroup, PO Box 100, 6700 AC Wageningen,The Netherlands Summary 1. This studyinvestigatedtheeffectof car traffic on thebreedingdensityofbirdsin deciduous and coniferouswoodland, and the importanceof noise and visibilityof cars as possible factorsaffecting density. 2. Of the 43 species analysed in both woodland types,26 species (60%) showed evidence of reduced densityadjacent to roads (based on analysiswithWilcoxon signed-rankstestand regression). 3. Regressionmodels withnoise load as the only independentvariable gave the best overall results.Calculated 'effectdistances'(the distancefromthe road up to where a reduced densitywas present)based on these regressionsvaried between speciesfrom40- 1500m fora road with10000 carsper day to 70-2800 m fora road with60000 carsper day (120 kmh-' and 70% amountofwoodland alongtheroad). For a zone of 250 m fromthe road the reductionof the densityvaried from20 to 98%. 4. Whenvisibility ofcarswas controlledfor,thenumberof speciesshowingdensity reductionswas muchhigheron plotswitha highnoise load thanon ones witha low noise load. When noise conditionswere held constant,however,there was no differencein bird densitiesbetweenplots withhighand low visibilityof cars. 5. It is argued that noise load is probably the most importantcause of the reduced densities.Visibilityof cars, directmortalityand pollutionare considered unimportant. 6. The resultsof thisstudystressthe importanceof consideringthe effectof car traffic on the breedingdensityof birdsin planningand constructing main roads. road planning Key-words:noise load, visibilityof cars, pollution,road mortality, and management. Journal of Applied Ecology (1995) 32, 187-202 Introduction 187 The effectof car traffic on birdshas receivedmuch attention,but most of the studies are focused on road victims(e.g. Hodson & Snow 1965; Bergmann 1974; Fullhaas et al. 1989; van den Tempel 1993). However, froma conservationpoint of view, it is much more interestingto know the consequences at the populationlevel. This has been poorlyinvestigated. The few available studies (woodland and * Present address: National ReferenceCentre for Nature,Forestand Landscape,PO Box 30, 6700 AA Wageningen, The Netherlands. open field)point out thatseveral territorial species show a lower breedingdensityin areas adjacent to roads thanin controlareas furtheraway (e.g. Clark & Karr 1979; Raty 1979;Ferris1979;van der Zande, ter Keurs & van der Weijden 1980; Adams & Geis 1981; Reijnen & Thissen 1987; Illner 1992a). In some species higherdensitieswere found adjacent to roads (Ferris1979; Clark & Karr 1979; Adams & Geis 1981), butthereis evidencethatthiswas due to different conditionsof the vegetationstructure close to the road. Estimatedeffectdistances(the distancefromthe road up to wherea lower densitycan be observed) 0047487 188 Main roads and densityof birdsin woodland Fig. 1. Main roads in The Netherlands(bold lines) and locationof paired plots in deciduous woodland (black dots) and coniferouswoodland (open dots). in open fieldhabitat extend to more than 1000m (van der Zande et al. 1980) and in woodland up to several hundredmetres(Rity 1979; Reijnen & Thissen 1987). Since there are indicationsthat the reductionofthedensityin the disturbedzone can be ratherlarge (Rdty 1979; van der Zande et al. 1980; Reijnen & Thissen 1987), this mightpoint to an importanteffecton breedingbird populations. How car trafficcauses densitiesto be reduced is unknown.It has been generallyassumed that the increaseof the mortalitydue to road traffic will be too small to cause a significantdecrease of the densityor the population'ssize (e.g. van der Zande etal. 1980; Leedy & Adams 1982; Ellenberg,Muller & Stottele1981). This impliesthat possible causes are probably related to emission of matter and energy by road traffic,such as pollution, visual stimuliand noise (van derZande etal. 1980; Reijnen & Thissen 1987). For woodland it is assumed that noise load, in particular,mightplay an important role (Reijnen & Thissen 1987; Reijnen & Foppen 1994). The objectiveofthisstudywas to determinewhich woodland species showed lower breedingdensities next to roads, over whichdistancessuch an effect occurred and to which degree the densitieswere lower. Furthermore, the assumptionwas testedthat noise is the mostcriticalcause of the effect. Methods STUDY AREAS AND SITE SELECTION The studywas carriedout in areas withdeciduous woodlandand coniferouswoodlandcrossedby main roads, scatteredover the Netherlands.The traffic densityof roads in areas withdeciduous woodland was 8000-61000 (mean 30334) cars per day and in areas with coniferouswoodland 29000-69000 (mean 45 319) cars per day. In deciduouswoodland 38 pairedplotsand in coniferouswoodland 17 paired plots were selected (Fig. 1). One plot of a pair was situatedadjacent to the road (road plot) and one plot at a distanceof on average400 m (controlplot). Withineach pair of plots habitat differenceswere controlledfor (see section of methods 'Control of 0047488 189 R. Reijnen et al. Table 1. Habitat characteristics of road and controlplotsin deciduousand coniferouswoodland (mean values). Statistical significance is based on Wilcoxonsigned-rankstest Deciduous woodland (n = 38) Coniferouswoodland (n = 17) Estimate Road Control Sign. Road Control Sign. Cover of vegetationlayers(%) <25 cm 25-50cm 50-100 cm 1-2 m 2-5 m 5-10m 10-20 m >20 m 59 31 30 19 42 55 49 14 55 35 28 21 41 51 53 20 NS NS NS NS NS NS NS 72 63 33 3 25 56 37 75 65 33 8 27 49 33 NS NS NS NS NS NS NS Treecharacteristics Circumference (cm) Distance between(m) Height (m) 87 3-1 16-4 86 2-8 17-0 NS NS NS 63 18-8 12-5 63 20-4 12-6 NS NS NS Size and edge Size (ha) Borderlinewithopen area (%) 4-04 53 4-24 53 NS NS 7-20 55 8-20 46 NS NS NS=P>G1G: ** P<OO1. other factors that influence breeding density'). Because available studiesin woodland indicatethat effectdistances can be rather large (Raty 1979; Reijnen& Thissen1987), some or partofthecontrol plots mightstill be influencedby car traffic.For practicalreasonsthesecontrolplotswerenotlocated fartheraway fromthe road. To obtainreliablebird densitiesa minimumplot area of 1-5ha was taken for deciduous woodland and 5 ha for coniferous woodland(based on Hustingsetal. 1985,see Table 1 formean size of plots). Investigationswere carried out in 1987 (deciduous woodland) and 1988 (coniferouswoodland). Oak (Quercus robur L.) was the dominanttree species in 30 of the 38 paired plots of deciduous woodland,poplar(Populus sp.) in five,beech (Fagus sylvaticaL.) in threeand alder (Alnus glutinosaL.) in two. The developmentand floristiccomposition of shruband herbaceouslayersalso variedbetween the paired plots. In all coniferouswoodland plots, pine (mostlyPinus sylvestris L.) was the dominant tree species and a shrub layer was almost absent. The herbaceous layer was characterizedby species of poor soil conditions, of which Deschampsia flexuosa(L.) Trin. was mostabundant. BREEDING ** DENSITY Breeding bird densities were measured as the numberofterritories perarea unitusingthemapping methodaccordingto Hustingset al. (1985). In the mapping method the number of territoriesof a species in an area is derived from all individual ofterritorial behaviourmade on several registrations visitsin the field. However, because of the rather small size of the plots, for some species that have large territories,the number of registrationswas used to calculate densities.To calculate the total densityof all species combined, the densities of species which were based on registrationswere dividedby two. Everyplot of a pair was visited12 timesfromthe beginningofMarchtilltheend ofJune.The distance of 50 m at which an observershould approach all partsof the plot (Hustingset al. 1985) was reduced to 25 m in order to minimizethe maskingeffectof trafficnoise on bird song. For territorieswhich overlappedthe boundariesof a plot, only the proportionof theterritory withinthe plot (based on the percentage of registrations)was included in the calculationof density. TRAFFIC LOAD To measurethe traffic load of plots or partof plots (see analysis)we used parametersforthe noise load and forthe visibilityof cars. Other possible causes, such as pollutionand road mortality,could not be measuredquantitatively. The noise load was estimatedusing an existing mathematicalmodel,whichexpressesthenoise level in dB(A) as the 24-hour value of the equivalent noise level (Moerkerken & Middendorp 1981). Calculated noise levels refer to points and the most relevantparametersin our studywere: traffic density,speed and type of cars, shortestdistance fromthe receiverpoint to the road and heightof road above groundlevel. Other parameterssuch as road surfaceand groundsurfacedid notvaryand for the receiverheight,1 m was taken (other heights 0047489 190 Main roads and densityof birdsin woodland 50 40 - 30 0 2 a 200 l l l|3 400 600 800 Distancefromthe road Cm) 1000 Fig. 2. Estimationof thenoise load ofcar traffic along a highwaywith50 000 cars per day and a speed of cars of 120kmhbased on Moerkerken& Middendorp(1981) and Huisman (1990) (see Methods forfurtherexplanation). 1 = open field; 2 = proportionof woodland along the road 05; 3 = proportionof woodland along the road 1 0. were investigated, but the calculatedvalues were all stronglycorrelatedwiththe 1 m data, r = >0-90). Because the model does not take into account the noise reducingeffectof woodland, which can be very important(Huisman 1990; Huisman & Attenborough1991), thecalculatednoiselevelswere adapted by usinga 'woodland effectterm'according to Huisman (1990). This 'woodland effectterm'is a functionof woodland type, the shortestdistance betweenroad and receiverpoint,thereceiverheight and the proportionof woodland between receiver pointand theroad (PW). It is specificforroad traffic noise and ranges from 0 to 12 dB(A) for most woodland types(deciduous and pine) and from0 to 16 dB(A) fordense coniferouswoodland.The maximum reductionis reached at 200 m fromthe road (see Fig. 2). Because changesin noise transmission between the road and the receiverpoint are most importantin a sightangle fromthereceiverpointto the road (Moerkerken& Middendorp1981), assessing of PW was restrictedto thissightangle (set at 143 degrees). To obtaina mean noise level fora plotor partof a plot, the value of one representativepoint was taken. This pointwas situatedin the middleof the line parallelto the road at a distanceof 10(log DL-Iog DS)/2 [log distanceis chosen,because noise levels in dB(A) have a logarithmic scale], in whichDL is the largestdistanceand DS the smallestdistanceof the plot fromthe road. Trafficdata were available in reportsof the Ministryof Traffic,Public Worksand Water Management(trafficdensity,speed of cars, type of cars). The distancefromthe representative pointto the road and PW were derivedfromcartographicmaps. The heightof the road above ground level and the woodland typewere estimatedin the field. of cars of plotsor partof To measurethevisibility plots,a simplemethodwas developed. The distance fromthe road where the trafficcould not be seen anymore,was thebasic estimate.Thiswas measured in early spring(leaves absent) and in late spring (leaves present). Assuming that the visibilityof cars decreases with the distance to the road, the relativevalue for a plot (between 0 = no visibility and 1 = visibilityin the verge of the road) was calculated as the proportionof the plot influenced by visibilitymultiplicatedby the mean level of the visibilityin the influencedpart of the plot (see Fig. 3). Differencesin traffic densityand typeofcars were not considered. It was assumed that when trafficis more or less continuous(>10000-20000 carsper day), thevisualexperienceoftraffic remains constant. CONTROL INFLUENCE OF OTHER BREEDING FACTORS THAT DENSITY the effectof road traffic, To identify we chose plots that were as similaras possible in other variables that influencebreeding density. Because of the paired design, the similarity was only requiredfor the plots of one pair. Variables taken into account werearea ofplots,vegetationstructure, surrounding landscape and management. Only plots along roads used by trafficfor more than 5 years were considered. To assess the similarityof vegetationstructure betweenthe road and controlplots of each pair the followingmeasurementsweremade. The vegetation structurewas measured along randomly chosen transectswitha lengthof 100m. To obtain values forthe vegetationcover, every5 m the presenceof vegetationwas scored at eightheightclasses. Also the distanceto the nearesttree fromthese points, and the heightand circumference of thistree were established.The proportionof the boundaryof the plot borderingopen fieldwas used to characterize the edge. In deciduouswoodland onlyone characteristic of the vegetation structure,the cover of the layer >20 m, showeda significant difference betweenroad and controlplots(Table 1). However,it is notlikely that this caused significant differencesin breeding 0047490 191 Studyplot R. Reijnenet al. 0-8 o 06 V -----*MV DL-DS*M MV 0-2 0- DS DM DN DL Distancefromthe road Fig. 3. Estimationof a relativevalue forthe visibilityof cars (V) in woodland (see also methods). DS= mean smallest distanceof plot to the road; DN = mean largestdistanceof plot to the road; DL = mean distancefromthe road where of traffic cannotbe seen anymore; DM = mean distancefromtheroad of thepartof theplot thatis influencedbyvisibility cars; MV = mean value forthe visibilityin the influencedpartof the plot (is value at DM). Because DS was alwaysvery in thevisibility ofcarsin thezone betweentheroadsideand DS were neglectedwithrespectto close to theroad, differences the visibilityparameter. densitiesbetweenroad and controlplots. The cover of the layer >20 m was of minor importancefor the structureof the deciduous woodlands, and the differencebetween road and controlplots was not in the heightof the trees. supportedby a difference There were no differencesof vegetationstructure betweenroad and controlplotsin coniferouswoodland plots. Since species were not always presentin all the pairedplotsof a series,additionaltestswere carried out on subsets of paired plots. The results were similaras compared with the whole dataset. Significantdifferences between road and controlplots wereonlypresentindeciduouswoodlandandrestricted to the cover of the layer>20 m. ANALYSIS In accordancewiththe studydesign(paired plots), the data were analysed at firstwith the Wilcoxon test.A non-parametric signed-ranks testwas chosen forrobustness.The basic analysiswas a comparison of densitiesbetweenroad and controlplots. Moreover, to detect possible effectson the densityat a veryshortdistancefromthe road, large road plots (at least twice the required minimumsize) were dividedintotwopartsthatwerenearto and farfrom the road. This was only done for road plots that had a very homogeneous vegetation structure. Because we expected lower densitiesclose to the in densitywere tested one-sided. road, differences Deciduous woodlandand coniferouswoodlandwere treated as separate series of paired plots and an analysisforindividualspecies was carriedout if the numberof territorieswas >10 or the number of registrations >20. In order to obtain informationabout the size of the effecton the densityand to determinethe relative importance of noise load and visibility of cars, a supplementaryanalysiswas carried out usinga regressionmodelwithnoiseload and visibility of cars as independentvariables. Since both measurementsof visibilityof cars were stronglycorrelated with each other (r = 0-95), only visibility of cars in early spring(withoutleaves) was used in the regressionanalysis.Althoughthe correlation between noise load and visibilityof cars was also high (r = 0 63), usingboth explanatoryvariablesin the regressionmodels was considered acceptable. For a more precise measurementof the effect distance,large homogeneousplotswere splitin two or sometimesmoresubplots(dependingon thesize). or As dependentvariable the numberof territories was taken and to model densities,the registrations logarithmof the area of the (sub) plotswas included in the model as an offset.Because counts tend to follow a Poisson-likedistributionin which the varianceis proportionalto themean, thenumbersof territories weremodelledby Poisson regressionwith correctionforpossibleoverdispersion(Jongman,ter Braak & van Tongeren 1987; see F-test below). Includinga factorforpairs in the model accounted betweenpairsof plots(whichmaybe fordifferences due to differencesin vegetationstructure,etc.). If load follows of birdsto traffic the level of sensitivity a Gaussian distribution,the relationshipbetween numbers and trafficload is sigmoidal. However, theoreticallythe most ideal combinationof fitting a sigmoidalrelationwitha Poisson regressionand a factor for pairs was very difficultto compute. Instead, a thresholdmodel was used: below the thresholdthereis no change in bird density;above the thresholdan equal percentageof the number of territoriesdisappears per unit of noise load. The thresholdvalue and the decrease factorwere estimated by log linear Poisson regressionwith 0047491 192 Main roads and densityof birdsin woodland simultaneousconfidenceregion for the threshold and the decrease factor. To get a furtherindicationof the importanceof noise load and visibilityof cars, subsets of paired plots were created in whichthe road plots differed A \ a) in one explanatoryvariable but not in the other (see Table 2). This was onlypossible fordeciduous woodland. If noise load is the mostimportantcause B of reduced densities, one would expect that the effect on the densityin the subsetwitha highnoise T R load is much largerthan in the subset with a low Noise load in dB(A) noise load. Between the subsets that differedin of thedecreasefactorof the density visibilityof cars, differencesin the effecton the Fig. 4. Estimation usinga threshold densityshouldthenbe muchsmalleror absent. The based on loglinearPoissonregression value in dB (A), model (see methods).T = threshold analysiswas done withthe Wilcoxon signed-ranks R = noiseload in dB (A) in thevergeof theroad. The test and regression. = areaofA/(areaofA + B). decreasefactor Because of thesmallsize of the data set, besides a level of level of P < 0*05,a significance a numericalsearch procedure for the best-fitting significance P < 0.10 was considered. We used a comparisonthresholdvalue. For the estimationof the decrease of the effectof the factorsee Fig. 4. The significance wise significancelevel, to avoid loss of power and 1989) noiselevelwas testedwiththeF-test(Jorgensen therebyaccept thatwe are likelyto marktoo many withtwo degreesof freedom(one forthe threshold species as showinga significance change. value and one forthe slope) in the numeratorand, therefore,had to be two-sided. However, in the Results absence of a threshold,the test was one-sided. Over-dispersionwas indicatedif the mean deviance DENSITY IN DECIDUOUS WOODLAND was largerthan 1. In such cases the mean deviance Of the 50 species presentin the 38 paired plots, 41 was used as denominatorof the F-statistic,other(see Table 3) weresufficiently wise the value 1 was used. An approximate90% numerousforanalysis. By usingthe Wilcoxonsigned-ranks test,the density confidenceinterval for the thresholdvalue was of 12 of the41 species was reducedclose to the road constructedas a by productof our numericalsearch thresholdvalue. The (at P < O10, six species at P O< 05, Table 3). For procedurefor the best-fitting three of these 12 species there was only an effect intervalcontainsthe thresholdvalues whichyielded on the densitywhen comparingparts of the road deviancevalues thatexceeded theminimumresidual deviance by no more than sF, wheres is the mean plots that were near to and far fromthe road (at P < 010, one species at p < 0-05, Table 3). Despite deviance(set to 1 ifs was smallerthan1) at thebestthe negativeeffecton severalspecies, the densityof fittingthresholdvalue and F is the 90% point of all species combinedwas not significantly an F-distributionwith one degree of freedom in reduced the numeratorand the degrees of freedomof the close to the road (Table 3). With regressionthe model with only noise load residual in the denominator.Because some plots reductionof the densityfor weresplitintosubplots,someindependencebetween resultedin a significant data was introduced.As a result,the confidence seven of the 12 species (at P < 0. 10, fivespecies at P < 0-05, Table 4) whichshowed an effectwiththe intervalfor the thresholdvalue may be somewhat paired test and also for four additionalspecies (at too narrow. We did not succeed in obtaining a Table 2. Trafficload of road plotsforsubsetsof paired plotsin deciduouswoodland (mean values). Statisticalsignificance is based on Wilcoxonsignedrankstest Subset of plots n Cars per day Noise load Visibilityof cars Noise load High Low 19 19 33179 27488 58 48 0-46 0 35 NS Visibilityof cars High Low 19 19 30562 30106 53 52 NS 0-61 0-21 NS=P>0 10; *** P<O 001. 0047492 193 R. Reijnen et al. Table 3. Density of breeding birds in road (R) and control (C) plots of deciduous woodland. Species for which a comparisonof densitieswithinroad plotswas carriedout are markedwithparenthesesand resultsare onlyshownifthey were significant' NT R C Sign. Species NP Anas platyrhynchos 10 16 0-10 0-05 NS Buteo buteo( 18 64 0-16 0-21 NS Phasanius colchicus 13 60 0-13 0-22 * Cuculus canorus 13 35 0-04 0-16 * Scolopaxrusticoka 11 42 0.11 0-09 NS Columba oenas ( Columbapalumbus 0 14 30 37 186 0-08 0-59 0-12 0-54 NS NS Picus viridus Dendrocopusminor Dendrocopusmajor 0 7 10 28 27 11 85 0-05 0-02 0-25 0-12 0-05 0-25 NS NS' 8 33 0-09 0-08 NS 0 troglodytus Troglodytus Prunellamodularis0 34 16 289 41 0-91 0-16 0-96 0-14 NS NS Erithacusrubecula0 Phoenicurusphoenicurus Turdusmerula( Turdusphilomelos() 34 10 38 24 357 17 350 112 1-09 0-05 1-15 0-29 1.00 0-06 1-04 0-32 NS NS NS NS Acrocephaluspalustris Hippolais icterina Sylviaborn ( Sylviaatricapilla0 Phylloscopussibilatrix Phylloscopuscollybita( Phylloscopustrochilus( Regulusregulus 5 7 28 35 10 35 26 7 11 16 175 172 22 222 264 23 0-04 0-11 0-70 0-56 0-04 0-87 0-70 0-05 0-15 0-15 0-82 0-62 0-10 0-90 0-98 0-07 t NS NS NS NS Muscicapa striata0 Ficedula hypoleuca0 26 13 85 27 0-27 0-09 0-24 0-06 NS NS Aegithaloscaudata 0 Parus montanus Parus palustris0 Parus caeruleus0 Parus major ( 25 18 23 33 37 39 31 65 252 504 0-17 0-09 0-17 0-79 1-63 0-13 0-19 0-20 0-75 1-45 NS NS NS NS NS Sittaeuropaea 0 0 Certhiabrachydactyla 18 30 65 142 0-22 0-44 0-17 0-44 NS NS' Oriolusoriolus 10 33 0-04 0-20 t Ganrulusglandarius( Pica pica ( Corvuscorone0 29 17 29 146 66 185 0-50 0-20 0-83 0-40 0-23 0-71 NS NS' NS Sturnusvulgaris0 20 121 0-41 0-45 NS Coccothraustes coccothraustes Fringillacoelebs ( 10 32 18 238 0-03 0-72 0-06 0-85 NS All species combined 38 4493 14 76 15-53 NS Anthustrivialis * * ** t t lower densityin partof road plots near to the road (N) thanin partof road plots furtheraway fromthe Significantly P <0-10 (N = 0 50, F = 0.66), Pica pica road (F). DendrocopusmajorP < 0-01 (N = 0-21,F = 0.41), Certhiabrachydactyla P<010 (N =0-33, F =0-56). For species in bold the data are or registrations. NP = numberof paired plotsinvolved;NT = totalnumberof territories test.NS = P > 0-10; t P < 0-10; is based on Wilcoxonsigned-ranks Statisticalsignificance based on numberof registrations. * P<0-05; ** P<0-01. 0047493 194 Main roads and densityof birdsin woodland ofcars (V) in deciduouswoodlandusinga Table 4. Significant regressionsofbreedingdensityon noise load (N) or visibility thresholdmodel (see Methods). For species in bold the data are based on numberof registrations Species N/V df Buteo buteo N 1/45 2-88 Phasanius colchicus N V 2/28 1/29 5 15 2-92 Scolopax rusticola N 2/28 Cuculus canorus N V Dendrocopusminor F T(CFL) C Sign. ?23 (?-50) -0 050 * 53 (45-58) 0 (?-7) -0 253 -0 009 * * 2-53 56 (53-62) -2-191 t 2/31 1/32 8-28 18-60 49 (40-52) 0 (?-24) -1 143 -0-064 ** N V 2/23 2/23 2-84 3-81 50 (43-56) 46 (33-59) -3 207 -1 191 t Phylloscopussibilatrix Phylloscopustrochilus N N V 1/24 2/61 2/61 5 41 4-30 4-30 ?26 (?-62) 39 (?-60) 62 (44-81) -0 052 -0 034 -0077 * * * Oriolusoriolus N V 1/27 1/27 14-0 13-0 ?23 (?-39) 0 (?-57) -0 121 -0 038 Pica pica N 2/43 2 95 Coccothraustes coccothraustes N 1/30 3 30 Fringillacoelebs N 1/66 All species combined N 1/81 * -0-086 t ?26 (?-56) -0-067 * 2 51 ?23 (?-60) -0-013 t 1 98 ?23 (?-60) -0 004 t 46 (40-56) df, degrees of freedom;F, change in F-statistic(see methods) afteradding noise load or visibilityof cars; T (CFL), thresholdvalue in dB (A) and 90% confidencelimits(? means thata lower limitcould not be calculated); C, regression coefficient. NS=P>O-10. t P<O-10. * P<005: ** P<O0O1; *** P<O0OO1. P<O*10, two species at P<O05, Table 4). Moreover,therewas an effecton thedensityof all species combined(at P < 0-10,Table 4). Addingvisibility of cars to the model and assuminginteractionbetween noise load and visibilityof cars, did not improve thesemodels. If noise load was replacedby visibility of cars, onlyfivespecies withsignificant regressions remained (at P < 0-05, Table 4) and the effecton the densityof all species combined disappeared. So the model with only noise load gives the best overallresults.For thefivespecies showingan effect withnoise load and visibilityof cars, the fitof the regressionmodels was similar.This means that,in these species, the effectcan eitherbe explainedby noise load or visibilityof cars. Effect distances (threshold value in m) were Table 5. Decrease factor(see Methods)of thebreedingdensityforspecies and all speciescombinedin deciduouswoodland in thezone betweentheroad and the thresholdvalue and in a fixedzone of 250 m fromtheroad. Based on regressionswith noise load as explanatoryvariable. For species in bold the data are based on numberof registrations Decrease factorup to 250m Species Decrease factor up to threshold 10000 cars per day 60000 carsper day Buteo buteo Phasanius colchicus Scolopax rusticola Cuculus canorus Dendrocopusminor Phylloscopussibilatrix Phylloscopustrochilus Oriolusorioles Pica pica Coccothraustes coccothraustes Fringillacoelebs 0 62 0 77 0 97 0 96 0 98 0 61 0-38 0 82 0 58 0 68 025 0 76 0 14 0 14 0 25 0 24 0 73 0 28 0.95 0 20 0 81 0 32 0 81 0 28 0 27 0 52 0 43 0 79 0 45 0 98 0 49 0-86 0 36 All species combined 0-09 0-11 0 13 0047494 S+ 195 R. Reijnen et al. S- Sco/opaxrust/co/C Phosonius co/chicus Dendrocopus minor Cuculus canorus Pica pica l Phyl/. troch//us [_ _ _ _ C. coccothroustes | Phy//.sibi/atrix Fringil/a coelebs _ _ _ _i/ _ ) _ _ _ _ _ Buteo buteo Oriolusori/ous All species combined 10 000 1000 100 10 Distance fromthe road (i) Fig. 5. Effectdistances(m) withconfidencelimitsin deciduouswoodland for10000 cars per day (------) and 60000 cars ). The car speed was fixedat 120kmh-1 and theamountofwoodlandalongtheroad at 70%. Effectdistances per day (---*--were derivedfromthe regressionwithnoise load as explanatoryvariable and a conversionof noise load to distancein m (see Methods). S = mean rangeof visibilityof cars withleaves (+) or withoutleaves (-). intensities estimatedfornoise load only.Two traffic were chosen thatcoveredthe whole rangeof intensitiespresentin the data set; 10000 and 60000 cars per day. Othervariableswhichinfluencedthe effect distancewere kept constant,such as speed of cars (120kmh-1) and amount of woodland along the road (70% = mean value for the plots). For 10000 cars per day the effectdistancesof species varied fromabout 40 to 1500m, and for60 000 cars per day from70 to 2800m (Fig. 5). However,theconfidence limitswere verylarge,especiallyforthe large effect distances.The decrease factorof the densityin the zone betweenthe road and the thresholdvalue was in most species veryhigh (median 0-69) (Table 5). Calculated fora fixedzone of 250m adjacent to the (10 000 road, the decrease factorwas stillsignificant cars per day, median 0-33; 60 000 cars per day, median 0.49) (Table 5). regressionwas carriedout withnoise load only. Of thefivespecies whichshowedan effect(at P < 0-10, two species at P < 0-05, Table 7), two also had an effectwiththe paired test. Again therewas also a clear effecton the densityof all species combined (at P<0-5, Table 7). Effectdistances and decrease factorswere calculated using the same trafficdata and amount of woodlandalongtheroad as fordeciduouswoodland. The effectdistances of species varied from50 to 790 m for10000 carsper day and from100 to 1750m for60 000 cars per day (Fig. 6). The decrease factor of the densityin the zone between the road and the thresholdvalue was in most species veryhigh (median 0.61) and remainedimportantif calculated fora fixedzone of 250 m (10 000 carsday-': median 0-23; 60000 carsday-': median 0-60) (Table 8). COMPARISON DENSITY IN CONIFEROUS WOODLAND Of the28 speciesoccurringin the 17 pairedplots,18 numerousto species (see Table 6) were sufficiently allow an analysis. By using the Wilcoxon signedrankstestonlyfourspeciesshoweda reduceddensity close to the road (at P < 0-10,3 species at P < 0-05, Table 6), but therewas a clear effecton the density of all species combined(at P < 0-05, Table 6). Because therewas a strongcorrelationbetween noise load and visibilityof cars (r = >0-90), the PAIRED OR PLOTS VISIBILITY DENSITIES OF DIFFERING OF IN IN SUBSETS NOISE OF LOAD CARS In the two subsets of paired plots that differedin noise load, but not in visibilityof cars, 28 species could be compared.WiththeWilcoxonsigned-ranks testninespecies showeda lowerdensityclose to the road in thesubsetwitha highnoise load (at P < 0-10, four at P < 0-05, Table 9) and only two in the subsetwitha low noise load (at P < 0-05, Table 9). Moreover, in the subset with a low noise load all 0047495 196 Main roads and densityof birdsin woodland Table6. Densityofbreeding birdsinroad(R) andcontrol (C) plotsof coniferous woodland.Speciesforwhicha comparison ofdensities within roadplotswascarriedout aremarkedwithparentheses andresults areonlyshownif theyweresignificant' Species NP NT R Columba palumbus () 11 48 Anthustrivialis ( 10 30 0-13 0-23 NS () Troglodytus troglodytes 14 37 0 16 0-22 NS Erithacusrubecula() Turdusmerula 16 13 55 25 0-32 0 26 NS 0-09 0-15 NS Sylviaatricapilla Phylloscopuscollybita() Phylloscopustrochilus Regulusregulus( 5 8 8 14 11 23 24 41 0.04 0-10 0 11 0 22 C Sign. 0-32 0-63 * of cars (at P<0-10, two at P<0-05, Table 11). In the regressionanalysis,withvisibilityof cars as the explanatoryvariable, the number of species that showed a negativeeffecton the densitywas two in plotswithhighvisibility(at P < 0-10, one species at P < 0-05,Table 12) and one in low visibility plots(at P < 0-10, Table 12). In none of these cases was an effecton the densityof all species combinedfound. Discussion REDUCTION OF DENSITIES Of the 43 species that could be analysed in the paired plotsof deciduousand coniferouswoodland, 26 species (60%) showed evidence of a reduced * density(Wilcoxonsigned-ranks testand regression). *l Althoughone can argue that thereis some uncertaintyforspecies thathad a weaklysignificant effect Parus montanus 12 19 0-06 0 11 NS (only P<0-10), there still remain 18 species with () Paruscristatus 16 48 0 29 0-33 NS Parus ater( 15 29 0-17 0-29 t a significanteffectat P < 0-05 (= 42%). This is Parus caeruleus 8 12 0 05 0 09 NS considerablymore than the few available studies Parus major () 17 49 0-26 0-27 NS indicate(Ferris 1979; Adams & Geis 1981; Reijnen & Thissen 1987). Nine of the 10 species which Garrulusglandarius 13 39 0-12 0-32 NS showed an effectin the studyof Reijnen & Thissen Pica pica ( 10 59 0-31 0-26 NS Corvuscorone( 11 108 0 42 0-38 NS (1987), also did so in this study. So, the results indicate that along busy roads, car trafficcausing Fringillacoelebs ( 13 77 0 45 0-52 NS reduceddensitiesofbreedingbirdsin woodland,is a generalphenomenon.This is reinforcedby the fact 0 17 256 2-35 3.39 * All speciescombined thatspeciesofverydifferent taxonomicgroupswere l Alsosignificant lowerdensity inpartofroadplotsnear affected. to theroad (N) thanin partof roadplotsfurther away Of the few available studies that give rough fromthe road (F). Regulus regulusP< 0.01 (N =0-04, estimatesof the effectdistanceand the size of the F= 0 22). effecton the density,only the studyof Reijnen & NP = number of paired plots involved, NT= total Thissen (1987) allows an appropriatecomparison number of territories or registrations. For speciesin bold thedata are basedon number of registrations. Statistical with our results of deciduous woodland. They significance is based on Wilcoxonsigned-ranks test. investigatedsimilar species along a highwaywith I NS=P>0*10; P<0 10; * P<0-05, ** P<0 01. 45 000 cars/dayand estimateda mean effectdistance of about 500 m and a decrease factorof the density species had an effectat a veryshortdistance(comfor all species combined of 0-15. Although the parisonof densitieswithinroad plots), whilein the decreasefactorforall speciescombinedis somewhat subset with a high noise load this was only the higherthanin thisstudy,the effectdistanceis much case in three of the nine species (Table 9). With lower. These differences could be due to the rough regression, using noise load as the explanatory and descriptivemethodused by Reijnen & Thissen variable, negativeeffectson the densitywere only (1987) (comparisonof densitiesin plots at different foundin the subsetof paired plotswitha highnoise distancesfromthe road). However, the difference load (seven species at P < 0-05, Table 10). This in the effectdistanceis not necessarilysignificant, supports the results obtained with the Wilcoxon sincethevalue of Reijnen & Thissen(1987) remains test.Furthermore, forthedensityof all signed-ranks withinour confidencelimits(see Fig. 5). Moreover, species combined,therewas only a negativeeffect our largesteffectdistancesmightbe overestimated, in the subset of paired plots witha highnoise load because they were obtained from regressionsin (at P < 0-05, Tables 9 and 10). whicha threshold was absent.The absenceof a thresFor the twosubsetsof pairedplotsthatdifferedin hold could be due to the relativelysmall numbers visibilityof cars but not in noise load, no clear of observationsat a greaterdistancefromthe road. patternarose. Here, 26 species could be compared. With the Wilcoxon signed-rankstest, only three POSSiBLE CAUSES OF REDUCED DENSITIES species showed a lowered densityclose to the road witha highvisibility of cars (at P < 0-10,one species The fact that significantrelationshipswere found at P < 0-05,Table 11) and threewitha low visibility betweentraffic relatedfactorsand breedingdensity 0 05 0.11 0 25 0 36 NS NS 0047496 197 R. Reijnenet al. S T: trog/odytus Anthus frivia/is Regulus regulus Porus coeruleus Phyll. trochl/us All species combined I I I I I'I I 1 10 I I ' I I I 1 11 100 I I I I I I I 1000 Distance from the road 10 000 (in) Fig. 6. Effectdistances(m) withconfidencelimitsin coniferouswoodland for10000 cars per day (---i---) and 60000 cars per day (---*---).The car speed was fixedat 120kmh-1 and theamountofwoodlandalongtheroad at 70%. Effectdistances were derivedfromthe regressionwithnoise load as explanatoryvariable and a conversionof noise load to distancein m (see Methods). S = mean rangeof visibilityof cars. suggestthat car trafficand not the presence of a road per se, is the main cause of the observed effects (see also van der Zande et al. 1980; Illner 1992a). This is supportedby the relativelysmall numbersof reduceddensitiesin the subsetof pairedplotswitha low noise load. This studysuggeststhatthe effectof car traffic on breedingbird densitiesin woodland can be largely explainedby noise load. For subsetsof paired plots whichdifferedin noise load, but not in visibilityof cars, a larger number of species had a reduced densitywith a high noise load than with a low noise load. On the otherhand, therewere no clear differencesbetween subsets of paired plots which differedin visibilityof cars but not in noise load. Furthermore,the few significant relationshipswith visibility,when noise load is held constant,are all for species in which noise levels produce models which are at least as good. Althoughthe absolute numbersof species involved are actuallynot very great, the noise load hypothesisis in particular supportedby the factthateffectson the densityof all species combinedwere onlyfoundin the subset of paired plotswitha highnoise load. Moreover,of all (sub)sets of paired plots, the numberof species thathad reduceddensitieswas highestin thissubset. Table 7. Significantregressionsof breedingdensityon noise load in coniferouswoodland using a thresholdmodel (see Methods) Species df F Anthustrivialis 2/23 4-12 Troglodytes troglodytes 2/25 2-72 Phylloscopustrochilus Regulusregulus 1/17 2/28 3-00 3-54 Parus caeruleus 1/19 2-46 All species combined 2/30 5-84 T(CFL) C Sign. 52 (47-58) -2 618 * 58 (46-59) -6-846 t ?29 (?-52) 44 (?-56) -0 043 -0-080 t ?29 (?-59) -0-057 t -0-042 ** 48 (38-56) * df,degreesof freedom;F, changein F-statistic(see methods)afteraddingnoise load; T(CFL), thresholdvalue in dB (A) and 90% confidencelimits(? meansthata lowerlimit could not be calculated); C, regressioncoefficient.t P<0-10; * P<0-05; ** P<0-01. 0047497 198 Main roads and densityof birdsin woodland Table 8. Decrease factor(see Methods) of the breeding densityfor species and all species combined in coniferous woodland in the zone between the road and the thresholdvalue and in a fixedzone of 250 fromthe road. Based on regressionswithnoise load as explanatoryvariable Decrease factorup to 250 m Species Decrease factor up to threshold 10000 cars per day 60 000 cars per day Anthustrivialis 0-98 0 20 0 39 Troglodytus troglodytes 0.99 0 11 0-23 Phylloscopustrochilus Regulusregulus 0-53 0-58 0-62 0-69 0 60 Parus caeruleus 0 61 0-23 0 78 All species combined 0-35 0 10 0 22 One can pointout thatthemeasurementof visibility of cars mightgive some uncertainty, because differences in traffic densitywere not takeninto account. However, it is unlikelythat this would be very important,since trafficdensitieswere ratherhigh and quite similarin all subsets of paired plots in deciduouswoodland (Table 2). Because other possible causes such as collisions and pollution(e.g. van derZande etal. 1980; Reijnen & Thissen1987) werenotconsideredin our analysis, thenoise load is notnecessarilytheonlycause of the reduceddensities.However,thereis some evidence that these other factorsare not veryimportant.It is assumed that an increase of mortalitydue to road traffic is notveryimportantin causingreduced densitiesnear the road (e.g. Leedy & Adams 1982; Ellenberget al. 1981). This is plausible for species whichhave small territories and thatare less likely to flyacross roads (+60% of the studied species), whichis supportedby a studyof Reijnen & Foppen (1994). They observed equal survivalrates of male willow warblers Phylloscopus trochilus (which showed a clear decreased densitynear the road in our plot series) close to a highwayand in areas at a distance of several hundredmetres. On the other hand, a studyby Sargeant(1981) indicatesthatalso in species withlarge territories (ducks), road mortality(0-2% of breedingpopulations)cannotbe very importantin reducing densities. Only for owls, in particularbarn owl Tyto alba, there is some indication that road mortalitymight influence population levels (van den Tempel 1993; Illner 1992b). However,owls were not presentin our data set. Pollutioncaused by road traffic can affectabundance and size ofinsectsadjacentto roads (Przybylski 1979; Bolsinger & Flickinger 1989) and therefore mighthave an effecton densitiesof breedingbirds by reducing the availabilityof insects as a food source. However,the reportedrangeof the effectis verysmall(up to 50 m fromtheroad in a ratheropen area). Moreover, of the 27 insectivorousspecies investigatedin thisstudy,only 16 species showed a reduced densityof which13 had a maximumeffect distancebetween 100 and 1500m. Pollutioncan also cause increasedlevels of toxic substancesin birds. This has been shown forlead, butthelevelswerefarbelow thetoxiclevel (Lowell, Best & Prather 1977; Grue et al. 1986), and an effecton reproductionand mortalitycould not be established(Lowell et al. 1977). There is also a more general indication that pollutionby car trafficis not very importantas a cause of reduceddensities.L.P. Kuggeleijn(personal communication)estimatedthat in woodland concentrationsof exhaust gases and other pollutants reach background levels within 50 m from a highwaywithdense traffic (40-50 000 cars per day). The majorityof the species in this study (75%), however,showedmaximumeffectdistancesbetween 100 and 1500m (see Figs 5 & 6). Very littleis knownabout how noise load could cause reduced densities of breeding birds. An obvious explanation would be disturbanceof the communicationpattern between birds (see also Martens, Foppen & Huisman 1985). Reijnen & Foppen (1994) found that male willow warblers Phylloscopustrochilus close to a highwayexperienced in attractingor keeping a female and difficulties moved away fromthe road in the followingyear. It was argued that distortionof the song mightbe a possiblecause ofthiseffect.In thehazelhenBonasia bonasia it was assumed that the decrease of the breedingdensityin partlyurbanizedareas was due to a more permanentincrease of the noise level, which increases predation by hamperinghearing alarm calls (Scherzinger 1973). However, there are some indicationsthat disturbanceof the communicationpatternis not a general mechanismin causingreduceddensities.Trafficnoise in woodland has a wide frequencyband of at least 100Hz to about 10 kHz withthe highestsound pressurelevels 0047498 199 R. Reijnenet al. Table 9. Densityof breedingbirdsin road (R) and control(C) plotsof deciduouswoodlandwithhighand low noise load. Species forwhicha comparisonof densitieswithinroad plotswas carriedout are markedwithparenthesesand resultsare onlyshowniftheywere significant (see footnote).NP = numberof paired plots involved.For species in bold the data are based on numberof registrations. Statisticalsignificance is based on Wilcoxonsigned-ranks test Noise load high Species NP R Noise load low C Sign. NP R C Sign. Phasanius colchicus 9 0 07 0-28 ** 5 019 0 17 NS Buteo buteo 9 0.15 0-12 NS 8 0-17 0-29 NS Scolopax rusticola 6 0-12 0 07 NS 5 0 11 0 11 NS Colomba oenas Columbapalumbus () 8 14 0 09 0 53 0-08 0 49 NS NS 6 16 0-06 0-62 0.15 0.55 NS NS Dendrocopusmajor () 11 022 0-25 NS' 17 0 28 0 25 NS' Troglodytus troglodytus( Prunellamodularis 17 9 0 70 0.19 0 78 0-18 NS NS 17 7 1 11 0-12 1-14 0-10 NS NS Erithacusrubecula( Turdusmerula() Turdusphilomelos( 16 19 13 1-15 1-07 040 1.11 0.99 0-41 NS NS NS' 18 19 10 1 00 1-24 0-20 0 90 1 10 0-25 NS NS NS Sylviaborn ( Sylviaatricapilla Phylloscopuscollybita() Phylloscopustrochilus() 14 17 18 13 0-62 0-48 0-78 0-69 1-13 0-72 094 1-12 * t t ** 14 17 17 13 0-78 0-65 096 0-72 0-51 0 57 0-86 0-83 NS NS NS NS Muscicapa striata() 13 0-26 0-24 NS 13 0-28 0-25 NS Aegithaloscaudata Parus montanus Parus palustris Parus caeruleus() Parus major () 12 8 11 17 19 0-14 0 10 0.20 0-78 1-58 0-16 0-14 0-25 0 74 1-54 NS NS f NS NS 13 10 12 16 18 0.19 0-08 0-14 0 79 1-67 0 10 0-23 0.15 0 75 1-36 NS NS NS NS NS Sittaeuropaea Certhiabrachydactyla 9 15 0-21 0-42 0-18 0-44 NS NS' 9 15 0-23 0 46 0-16 0 44 NS NS' Garrulusglandarius0 Pica pica ( Corvuscorone 15 8 14 053 0-15 0-82 0-42 0-16 0-46 NS NS NS 14 9 15 0 47 0 24 0-84 0 38 0 30 0 97 NS NS NS Sturnusvulgaris 10 0 45 0.59 NS 10 0-36 0-31 NS Fringillacoelebs 16 0-61 0-87 NS 16 0-82 0-83 NS All species combined0 19 13 3 15 8 19 14 1 13 3 NS 1 Significant lowerdensityin partof road plotsnear to theroad (N) thanin partof road plotsfurther away fromtheroad (F). Dendrocopusmajor P < 0-05 (noise load high,N = 0 10, F = 0.25) and P < 0.01 (noise load low, N = 0-17, F = 0.45), TurdusphilomelosP < 0 10 (N = 0-49, F = 0 79), Certhiabrachydactyla P < 0 10 (noise load high,N = 0 44, F = 0.64) and P < 0 05 (noise load low, N= 057, F = 0*83). NS = P> 0 10; U P< 0-10; * P< 005; ** P< 0-01. between 100-200Hz and 0 5-4kHz (Huisman & Attenborough1991). So, if disturbanceof communication is an important mechanism, those species producingsongs and calls witha frequency band similarto these highestsound pressurelevels oftraffic noisewouldbe expectedto be morevulnerable. Frequencybandsofsongand calls ofbirdswere taken from Bergmann & Helb (1982). However, there was no clear pattern(chi-squaredtest, 2df, P>O.10, Table 13), but the numbersare low and hardlyallow statisticaltesting.Because there are indicationsthat noise load due to car trafficcan cause stress in birds (Helb & Huppop 1991), an alternativeor, more probably, a supplementary explanationcould be thatbirdsavoid road habitats because of thisaspect (cf. Reijnen & Foppen 1994). 0047499 200 Main roads and densityof birdsin woodland Table 10. Significantregressionof breedingdensityon noise load in deciduous woodland withhigh noise load using a thresholdmodel (see Methods). For species in bold the data are based on numberof registrations Species df F T (CFL) C Sign. Phasanius colchicus Scolopax rusticola Sylviaborin Sylviaatricapilla Phylloscopustrochilus Pica pica Fringillacoelebs All species combined 2/15 2/11 1/27 1/37 2/29 2/22 1/33 1/38 5-28 3 99 3-24 4-20 5-73 3-45 3-39 5 30 40 (?-58) 56 (55-61) ?23 (?-54) ?23 (?-48) 38 (?-50) 46 (33-56) ?23 (?-56) ?23 (?-48) -0 099 -2-515 -0-024 -0-036 -0-046 -0 092 -0 019 -0-008 * * * * ** * * * df, degrees of freedom;F, change in F-statistic(see methods) afteradding noise load; T(CFL), thresholdvalue in dB(A) and 90% confidencelimits(? means that a * P < 0 05; ** P < 0.01. lower limitcould not be calculated); C, regressioncoefficient. of Table 11. Densityof breedingbirdsin road (R) and control(C) plotsof deciduouswoodlandwithhighand low visibility cars. Species forwhicha comparisonof densitieswithinroad plotswas carriedout are markedwithparenthesesand results l are only shownif theywere significant Visibilityof cars low Visibilityof cars high Species NP R C Sign. NP R C Sign. Phasanius colchicus 5 0-21 0-24 NS 8 0 05 0-20 ** Buteo buteo 9 0-18 0-25 NS 8 0 13 0-16 NS Columbapalumbus () 15 055 0 59 NS 15 0-63 0-49 NS Dendrocopusmajor () 16 032 0-28 NS 12 0-17 0-21 NS ( troglodytes Troglodytus Prunellamodularis 17 8 0-92 0-16 0-96 0-13 NS NS 17 8 0-89 0-15 0 90 0-16 NS NS Erithacusrubecula() Turdusmerula( Turdusphilomelos() 20 19 15 1-21 1-36 026 1-26 1-10 0 33 NS NS NS 15 19 9 0-76 0 94 0-33 0 81 0-98 0-32 NS NS NS Sylviaborn () Sylviaatricapilla( Phylloscopuscollybita() Phylloscopustrochilus() 13 18 17 12 0 54 050 0 74 026 0-48 061 0-85 0-66 NS tI NS **1 15 16 18 14 0-86 0-63 1-00 1-14 1-16 0-67 0-96 140 NS NS NS t Muscicapa striata() 14 0 35 0-28 NS 13 0 41 0-41 NS Aegithaloscaudata Parus palustris Parus montanus Parus caeruleus( Parus major ( 13 15 7 18 19 0-18 0-25 0-08 0 94 1-83 0-14 0-32 0-13 0-86 1 58 NS NS NS NS NS 12 8 11 15 19 0-15 0 10 0-10 0-63 1-28 0-12 0-08 0-25 0-64 1-28 NS NS NS NS NS Sittaeuropaea Certhiabrachydactyla ( 11 18 0-35 053 0-22 0-62 NS NS' 7 12 0 09 0 35 0-12 0-26 NS NS Garrulusglandarius0 Pica pica ( Corvuscorone0 15 7 14 0-51 0-16 0 54 0-42 0-15 0 59 NS NS NS 14 10 15 0*50 0-23 1-07 0-38 0-31 0-83 NS NS NS Sturnusvulgaris 14 0-68 0-67 NS 6 0-14 0-22 NS Fringillacoelebs 17 0-82 0 97 NS 15 0-61 0 73 NS All species combined 0 15 13-7 NS 16 12-5 NS 13-0 11-7 away fromthe road lowerdensityin partof road plotsnear to theroad (N) thanin partof road plotsfurther lSignificant Phylloscopus Sylvia atricapillaP <0-10 (N =0-45, F =063), (F). Dendrocopus major P K0*01 (N =0-18, F =045), P <0-10 (N = 0.41, F =0.64) . brachydactyla P <0 05 (N = 0-.17,F =0.84), Certhia trochilus NP = numberof paired plots involved.For species in bold the data are based on numberof registrations.Statistical P< 010; * P<0 005; ** P< 0.01. is based on Wilcoxonsigned-rankstest. NS =P> 010; significance 0047500 201 R. Reijnenet al. of cars in deciduouswoodlandwithhigh(H) and low (L) regressionsof breedingdensityon visibility Table 12. Significant of cars usinga thresholdmodel (see Methods) visibility Species H/L DF F T (CFL) C Sign. Phasanius colchicus Phylloscopustrochilus Fringillacoelebs H H L 2/7 1/27 1/33 3-38 12-16 2-74 50 (11-71) 0 (?-41) 0 (?-57) -0-058 -0-017 -0 01 t ** t of cars; T (CFL), thresholdvalue in df,degreesof freedom;F, changein F-statistic(see Methods) afteraddingvisibility dB(A) and 90% confidencelimits(? meansthata lowerlimitcould notbe calculated); C, regressioncoefficient.I P < 0 10; ** P < 0.01. Table 13. Reductionof densityof species relatedto coverageof thefrequencyspectrumof song and calls by thefrequency noise (see also Discussion) spectrumof traffic Numberof species with POPULATION PRACTICAL Frequencyspectrumof traffic noise, and of song and calls of the studiedbird species Reduced density No effecton density Similar Partlysimilar Not similar 9 7 10 6 7 4 CONSEQUENCES AND IMPLICATIONS Because the analysiswas based on the hypothesis that there is a negative relationshipbetween car and breedingdensity,possiblepositiverelationtraffic ships could not be detected. One mightargue that species which are not affectedby road trafficcan reach higherdensitiesifrelatedspecies are affected (due to competition).The presenceof road victims could also favour feeding conditions of carrion eaters,such as manycorvidspecies (e.g. Ellenberg et al. 1981). This would diminishthe overall effect on the breeding density.However, such patterns could not be detected in our data. One corvid species,themagpiePica pica, even showeda negative effecton the breedingdensity.Moreover, in both woodland types the total density of all species combinedwas lowered close to the road. The fact thatin deciduouswoodland onlya weak significance was found, could be due to the large number of paired plotswitha low noise load close to the road. For the subset of paired plots with a high noise load the densityof all species combinedwas clearly reduced (Tables 9 & 10). The available studiesalso do not indicatepositive Althoughhigherdensitiesclose effectsof car traffic. to roads were found in some studies (Ferris 1979; Clark & Karr 1979; Adams & Geis 1981), in thefirst in the two studiesthiswas explainedby differences vegetationstructure.For the study of Adams & Geis (1981) thiswas shownby a roughrecalculation of the data. They investigatedbreeding densities of species in transects perpendicular to roads, and observed lower densitiesclose to the road in seven species and higherdensitiesin nine species. Excludingthe partof the transectsclose to the road (which had differenthabitat features),six species had lowered densitiestowardsthe road and higher densitiestowardstheroad were no longerobserved. Since road trafficaffectsa large proportionof breedingbird species in woodland and causes significantreductionsin densityup to large distances fromthe road, the effectis probablyimportantin breedingbird populationsin largerareas. affecting This is reinforcedby the assumptionthat lower densitiesindicate a lower habitat quality (van der Zande et al. 1980; Reijnen & Thissen 1987), which was verifiedfor the willow warbler (Reijnen & Foppen 1994). Because differencesin densitywill in quality(e.g not alwaysequally reflectdifferences van Horne 1983;Bernstein,Krebs& Kacelnik1991), thereductionof habitatqualityalong roads mightbe underestimated. The resultsof thisstudystressthe importanceof on the breeding consideringthe effectof car traffic densityof birds in designingnew and modifying existingmajor roads. If noise load is the most importantcause of reduced densities,it is possible noise-muffling to reducetheseeffectsbyconstructing devices along the roads. However, to justifysuch evidence itis necessaryto providefurther expenditure that noise load is the main cause of the reduced densities of many bird species adjacent to busy roads. Acknowledgements We thank Carl Derks, Ton Eggenhuizen, Henk Meeuwsen and Jaap Mekel for theirassistancein the field,S.B. den Ouden and Henk Meeuwsen for preparingthe data, the manylandownersforkindly us to work on theirproperty,and Paul permitting Opdam, Helias Udo de Haes and two anonymous 0047501 202 Main roads and densityof birdsin woodland refereesfor their many valuable remarkson the manuscript.The studywas supportedby the Road and HydraulicEngineeringDivision of the Ministry ofTransport,PublicWorksand WaterManagement. References Adams, L.W. & Geis, A.D. (1981) Effectsof Highwayson Wildlife. Report No. FHWA/RD-81/067, Office of US DeparResearch,Federal HighwayAdministration, Washington. tmentof Administration, Bergmann,H-H. (1974) Zur Phanologie und Oekologie des Strassentodesder Vogel. Vogelwelt,95, 1-21. Bergmann,H.H. & Helb, H-W. (1982) Stimmender Vogel. Munchen. BLV Verlagsgesellschaft, Bernstein,C., Krebs, J.R. & Kacelnik, A. (1991) Distributionof birdsamongsthabitats:theoryand relevance to conservation.Bird Population Studies,Relevance to Conservationand Management(eds C.M. Perrins,J-D. Lebreton & G.J.M. Hirons), pp. 317-345. Oxford UniversityPress, Oxford. Bolsinger,M. & Fluckinger,W. (1989) Ambiantair pollutioninducedchangesin amino acid patternof phloem sap in hostplants - relevance to aphid infestation. pollution,56, 209-216. Environmental Clark, W.D. & Karr, J.R. (1979) Effectsof highwayson red-wingedblackbird and horned lark populations. WilsonBulletin,91, 143-145. Ellenberg,H., Muller, K. & Stottele,T. (1981) StrassenOkologie. Okologie und Strasse, pp. 19-115. Broschurenreihede Deutschen Strassenliga,Bonn. Ferris, C.R. (1979) Effectsof Interstate95 on breeding Management, Maine. Journalof Wildlife birdsin northern 43, 421-427. Fullhaas, U., Klemp, C., Kordes, A., Ottersberg,H., Pirmann,M., Thiessen,A., Tschoetschel,C. & Zucchi, zum Strassentodvon Vogeln, H. (1989) Untersuchungen Saugetieren, Amphibien und Reptilien. Beitrage zur NaturkundeNiedersachens,42, 129-147. Grue, C.E., Hoffman,D.J., Nelson Beyer,W. & Franson, L.P. (1986) Lead concentrationsand reproductive success in European starlings(Sturnusvulgaris)nesting Pollution withinhighwayroadsideverges.Environmental (SeriesA), 42, 157-182. Helb, H-W. & Huppop, 0. (1991) Herzschlagrateals Mass zurBeurteilungdes Einflussesvon Storungenbei Vogeln. 1992 (eds Hrsg. E. Bezzel, P.H. Ornithologenkalender Barthel,H.-H. Bergmann,H.-W. Helb & K. Witt),pp. 217-230. Aula-Verlag,Wiesbaden. Hodson, N.L. & Snow, D.W. (1965) The road deaths enquiry,1960-61. Bird Study,12, 90-99. Horne, B. van (1983) Densityas a misleadingindicatorof habitat quality. Journal of WildlifeManagement,47, 893-901. Huisman, W. (1990) Sound propagationover vegetationof Nijmegen, coveredground.Thesis,CatholicUniversity Nijmegen. Huisman,W.T. & Attenborough,K. (1991) Reverberation and attenuationin a pine forest. JournalAcoustical Societyof America,90, 2664-2677. Hustings, M.F.H., Kwak, R.G.M., Opdam, P.F.M. & achtergronReijnen,M.J.S.M. (1985) Vogelifientarisatie: Pudoc, Wageningen. en verslaglegging. den, richtlijnen on grey Illner,H. (1992a) Effectof roads withheavytraffic partridgedensity(Perdixperdix) density.Gibier Faune Sauvage, 9, 467-480. Illner, H. (1992b) Road deaths of Westphalian owls: methodologicalproblems, influenceof road type and possible effectson populationlevels. The Ecology and Conservationof European Owls (eds C.A. Galbraith, I.R. Taylor & S. Percival), pp. 94-100. JointNature ConservationCommittee,Peterborough. Jongman,R.H.G., Braak, C.J.F. ter& Tongeren,O.F.R. and Landscape van (1987) Data Analysisin Community Ecology. Pudoc, Wageningen. Jorgensen,B. (1989) Maximumlikehood estimationand large-sampleinferencefor generalizedlinear and nonlinearregressionmodels. Biometrika,70, 19-28. Considerations Leedy, D.L. & Adams, L.W. (1982) Wildlife in Planningand Managing HighwayCorridors.Report No. FWHA-TS-82-212, Office of Research, Federal Highway Administration,US Department of Administration,Washington. Lowell, L.G., Best, L.B. & Prather,M. (1977) Lead in urbanand ruralsongbirds.Environmental Pollution,12, 235-238. Martens,M.J.M., Foppen, R.P.B. & Huisman, W.H.T. (1985) Wordtvogelgeluidbeinvloeddoor het akoestisch milieu?Het Vogeljaar,33, 110-120. Moerkerken,A. & Middendorp,A.G.M. (1981) Berekening van wegverkeersgeluid. 's-Gravenhage. Staatsuitgeverij, Przybylski,Z. (1979) The effectsof automobile exhaust gases on the arthropodsof cultivatedplants,meadows and orchards.Environmental Pollution,19, 157-161. Reijnen, R. & Foppen, R. (1994) The effectsof car traffic on breedingbird populationsin woodland. I. Evidence of reduced habitat quality for willow warblers (Phylloscopus trochilus)breedingclose to a highway.Journal of Applied Ecology, 31, 85-94. Reijnen, M.J.S.M. & Thissen, J.B.M. (1987) The effects fromroad traffic on breeding-bird populationsin woodland. Annual Report 1986, pp. 121-132. Research InstituteforNature Management,Leersum. Raty, M. (1979) Effectof highwaytrafficon tetraonid densities.Ornis Fennica,56, 169-170. Sargeant,A.B. (1981) Road casualties of prairienesting ducks. WildlifeSocietyBulletin,9, 65-69. Scherzinger,W. (1979) Zum Feindverhaltendes HaselhuhnesBonasia bonasia. Die Vogelwelt;100, 205-217. door het Tempel, R. van den (1993) Vogelslachtoffers Nederland,Zeist. Vogelbescherming wegverkeer. Zande, A.N. van der, Keurs, W.J. ter & Weijden, W.J. van der (1980) The impactof roads on the densitiesof fourbirdspecies in an open fieldhabitat- evidence of a long-distanceeffect. Biological Conservation, 18, 299-321. Received26 October1993; revisionreceived1 March 1994