+ All Categories
Home > Documents > Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

Date post: 08-Dec-2016
Category:
Upload: susanna
View: 232 times
Download: 7 times
Share this document with a friend
9
Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines Felix Liechti , Jérôme Guélat, Susanna Komenda-Zehnder Swiss Ornithological Institute, 6204 Sempach, Switzerland article info Article history: Received 8 September 2012 Received in revised form 7 March 2013 Accepted 10 March 2013 Keywords: Migration model Passerine migrants Collision risk Wind turbines Spatial modelling Switzerland abstract Bird migration and wind energy production exploit a similar airspace. There is a well-founded claim that conservation should aim at facilitating both activities. Negative effects can be mitigated either by avoid- ing a spatial concurrence or by accounting for the temporal course of migration and stopping wind tur- bines during peak flight activities. In this study we promote a new methodological approach to reduce potential conflicts in the planning as well as during the operation phase of a wind energy project. The basis is a new spatially explicit model for broad front migration. It allows to quantify the collision risk with respect to topography. We simulated migration of non-soaring birds across Switzerland. Model parameters were tuned to achieve results in accordance with current expert knowledge based on many years of visual observations and radar measurements. The resulting maps were used to define areas with different collision risks. For medium and high risk areas, we propose a permanent monitoring system, which is able to shut down the local turbines during peak migration. We evaluated the impact of such a shutdown regime in five specific sites with quantitative radar data for at least one migration season. The model presented here is a simple preliminary, but robust, approach. The main weakness of the model is the use of large-scale rather than local wind conditions. Within the Alps, local wind fields can differ considerably from the general pattern, and accordingly also the distribution of flight directions. We hope to provide a basis for similar models in other geographic areas. In addition, we call for the use of large scale monitoring data, as hopefully will soon be available from weather radar networks, to validate any kind of spatially explicit migration models. Ó 2013 Elsevier Ltd. All rights reserved. 1. Introduction Both bird migration and wind energy production exploit the air- space in their own specific spatial and temporal pattern. Bird migration is a worldwide phenomenon across land and sea (Aler- stam, 1990; Newton, 2008). More than 20 billion birds regularly move twice a year between their breeding and non-breeding ranges. Distances covered can be tens to thousands of kilometres and flight altitudes range from a few metres above ground to thou- sands of metres in the free airspace. Wind is widely considered as a sustainable source of energy, and hence, wind farms are initiated rapidly worldwide (Global Wind Report, 2011). Apart from the po- sitive effect of renewable energy production, wind turbines can also have negative effects on the natural environment. One of them is the collision risk of migrant birds with wind turbines reaching up to 200 m above ground (e.g. Drewitt and Langston, 2006; Hüp- pop et al., 2006). The best way to mitigate conflicts between birds and wind tur- bines is to avoid their spatial concurrence (Bright et al., 2008). The distribution of breeding birds is often well known or can be estab- lished with an environmental impact assessment in due time. In contrast, the movement patterns of migratory birds are still poorly known, and because of the strong influence of weather, the tempo- ral and spatial patterns may differ considerably across seasons and years. Only very few bird species migrate within a relatively nar- row corridor (e.g. cranes Grus grus), but even the flight routes of such species vary substantially between years. Most migratory bird species move between breeding and non-breeding grounds in a broad front affected only by wind and topography. Migratory birds usually focus along topographical features like coastlines, straits, mountain ranges and passes (e.g. Kerlinger, 1989; Bruderer and Jenni, 1990; Bruderer and Liechti, 1999). In addition, bird migra- tion is not only concentrated in space, but also in time. Weather conditions have a strong influence on the timing and can cause mass migration restricted to a few days within a much longer migratory season (e.g. Erni et al., 2002; Van Belle et al., 2007). Therefore, measures to reduce the impact of wind energy produc- tion on migratory birds include avoiding high concentration areas by wind farms or closing down wind turbines when migration oc- curs at a high intensity. To apply such measures, the spatial and temporal distribution of migratory birds needs to be known. 0006-3207/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biocon.2013.03.018 Corresponding author. Tel.: +41 414629782; fax: +41 414629710. E-mail address: [email protected] (F. Liechti). Biological Conservation 162 (2013) 24–32 Contents lists available at SciVerse ScienceDirect Biological Conservation journal homepage: www.elsevier.com/locate/biocon
Transcript
Page 1: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

Biological Conservation 162 (2013) 24–32

Contents lists available at SciVerse ScienceDirect

Biological Conservation

journal homepage: www.elsevier .com/locate /b iocon

Modelling the spatial concentrations of bird migration to assess conflictswith wind turbines

0006-3207/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.biocon.2013.03.018

⇑ Corresponding author. Tel.: +41 414629782; fax: +41 414629710.E-mail address: [email protected] (F. Liechti).

Felix Liechti ⇑, Jérôme Guélat, Susanna Komenda-ZehnderSwiss Ornithological Institute, 6204 Sempach, Switzerland

a r t i c l e i n f o

Article history:Received 8 September 2012Received in revised form 7 March 2013Accepted 10 March 2013

Keywords:Migration modelPasserine migrantsCollision riskWind turbinesSpatial modellingSwitzerland

a b s t r a c t

Bird migration and wind energy production exploit a similar airspace. There is a well-founded claim thatconservation should aim at facilitating both activities. Negative effects can be mitigated either by avoid-ing a spatial concurrence or by accounting for the temporal course of migration and stopping wind tur-bines during peak flight activities. In this study we promote a new methodological approach to reducepotential conflicts in the planning as well as during the operation phase of a wind energy project. Thebasis is a new spatially explicit model for broad front migration. It allows to quantify the collision riskwith respect to topography. We simulated migration of non-soaring birds across Switzerland. Modelparameters were tuned to achieve results in accordance with current expert knowledge based on manyyears of visual observations and radar measurements. The resulting maps were used to define areas withdifferent collision risks. For medium and high risk areas, we propose a permanent monitoring system,which is able to shut down the local turbines during peak migration. We evaluated the impact of sucha shutdown regime in five specific sites with quantitative radar data for at least one migration season.The model presented here is a simple preliminary, but robust, approach. The main weakness of the modelis the use of large-scale rather than local wind conditions. Within the Alps, local wind fields can differconsiderably from the general pattern, and accordingly also the distribution of flight directions. We hopeto provide a basis for similar models in other geographic areas. In addition, we call for the use of largescale monitoring data, as hopefully will soon be available from weather radar networks, to validateany kind of spatially explicit migration models.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Both bird migration and wind energy production exploit the air-space in their own specific spatial and temporal pattern. Birdmigration is a worldwide phenomenon across land and sea (Aler-stam, 1990; Newton, 2008). More than 20 billion birds regularlymove twice a year between their breeding and non-breedingranges. Distances covered can be tens to thousands of kilometresand flight altitudes range from a few metres above ground to thou-sands of metres in the free airspace. Wind is widely considered as asustainable source of energy, and hence, wind farms are initiatedrapidly worldwide (Global Wind Report, 2011). Apart from the po-sitive effect of renewable energy production, wind turbines canalso have negative effects on the natural environment. One of themis the collision risk of migrant birds with wind turbines reachingup to 200 m above ground (e.g. Drewitt and Langston, 2006; Hüp-pop et al., 2006).

The best way to mitigate conflicts between birds and wind tur-bines is to avoid their spatial concurrence (Bright et al., 2008). The

distribution of breeding birds is often well known or can be estab-lished with an environmental impact assessment in due time. Incontrast, the movement patterns of migratory birds are still poorlyknown, and because of the strong influence of weather, the tempo-ral and spatial patterns may differ considerably across seasons andyears. Only very few bird species migrate within a relatively nar-row corridor (e.g. cranes Grus grus), but even the flight routes ofsuch species vary substantially between years. Most migratory birdspecies move between breeding and non-breeding grounds in abroad front affected only by wind and topography. Migratory birdsusually focus along topographical features like coastlines, straits,mountain ranges and passes (e.g. Kerlinger, 1989; Bruderer andJenni, 1990; Bruderer and Liechti, 1999). In addition, bird migra-tion is not only concentrated in space, but also in time. Weatherconditions have a strong influence on the timing and can causemass migration restricted to a few days within a much longermigratory season (e.g. Erni et al., 2002; Van Belle et al., 2007).Therefore, measures to reduce the impact of wind energy produc-tion on migratory birds include avoiding high concentration areasby wind farms or closing down wind turbines when migration oc-curs at a high intensity. To apply such measures, the spatial andtemporal distribution of migratory birds needs to be known.

Page 2: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

F. Liechti et al. / Biological Conservation 162 (2013) 24–32 25

The overarching goal of this study was to develop a novel meth-odology. We present a new approach to model the general move-ment and local concentration of broad front bird migration in astrongly structured topography. The aim was to quantify the spa-tial distribution of the potential collision risk for migratory birdsby modelling the spatial pattern of broad front bird migration forthe whole of Switzerland. The main challenge was to predict theintensity of bird migration within 200 m above ground level (agl).

There are two distinctly different flight behaviours for migra-tory birds (Pennycuick, 1975). To save energy, large species mainlysoar and glide. Since updrafts are site and weather specific, thisflight type is restricted in space and time. Small birds, such aspasserines, use powered flight, which, in principle, allows themto travel across any terrain at any time. About two thirds of thesespecies migrate at night across Switzerland (Winkler, 1999). Werestricted the model to species using powered flight, which makeup the vast majority of migratory birds. From the results, we de-rived a sensitivity map for the potential collision rates of migratingbirds with wind turbines. We also evaluated the mitigating effectof a stop and go regime for wind turbines driven by the intensityof bird migration based on real radar observations. We designatewhere a stop and go regime would be essential to keep the annualpotential collision rate below a given threshold. Since bird strikescan never be ruled out completely, a maximum number of accept-able fatalities is discussed.

2. Material and methods

2.1. The model

The study area included the Swiss territory and some regionsnearby (Fig. 1). The topography is characterised by three basicareas: the Alps in the south, the Jura in the northwest and the pla-teau lying in between. Many mountain tops of the Alps are around4000 m above sea level (asl), the peaks of the Jura are around1700 m asl. The plateau lies between 400 and 700 m asl.

We used a two-dimensional cellular automaton with absorbingedges to simulate broad front migration across Switzerland (Pack-ard and Wolfram, 1985). The area for simulation ranged from48.0�N to 45.6�N and 5.2�E to 12.9�E. The grid consisted of 260rows and 592 columns, with a 1 km resolution. To simulate au-tumn migration, we used an east–west starting line 20 km northof the study area, extending 60 km westwards and 185 km east-wards of the Swiss territory. For spring migration, the starting line

Fig. 1. Topography of Switzerland and nearby regions. The Swiss lowland extendsfrom the Lake of Constance in the NE to the Lake of Geneva in the SW. The mountainranges of the Alps cover the southern half of Switzerland, and the lower mountainsof the Jura arise along the north-western border. Indicated are the capital (Berne)and the five sites, where radar observations were carried out (copyright institute ofcartography and geoinformation, ETH).

was set on an east–west line, 20 km south of the study area and ona north–south line 60 km to the west. The simulation started withan equal number of birds in the cells of the starting lines, except forthose cells in the south, within the main mountain range of theAlps (cells 60–200). These grid cells were filled with half as manybirds as the other cells to account for the expected lower numberof migrants stopping over within the Alps and thus entering thestudy area from this region. Iteratively, a density probability wascalculated for each cell based on the densities in the eight sur-rounding cells. Each surrounding cell contributed a certain proba-bility and a flight altitude, which then for the focal cell wassummarized, or averaged, respectively. The density probabilitiesto move from one cell to a neighbouring cell were calculated stepby step, according to the preferred flight directions of the birds,and according to the height differences between the elevation ofthe surrounding cells and the mean flight altitude within the cellof origin (see below and Supplementary material Figs. A1 and A2).

2.2. Flight directions

The flight directions were implemented as the probability ofmoving from the cell of origin to one of the eight neighbouringgrid-cells. We used the results from radar observations in southernGermany and Switzerland (Bruderer and Liechti, 1990) to define amean and scatter of preferred flight directions under various windconditions. In Central Europe, migrant birds show a directionalpreference according to the season, i.e. 225� in autumn and 45�in spring. Weak winds (<5 m s�1) alter the distribution only mar-ginally and indifferently of wind directions (Liechti, 1993). Tailwinds narrow the range of flight directions, whereas head windsincrease the scatter (Bruderer and Liechti, 1990). The particularSwiss topography dominated by the Jura and the Alps and thebroader European weather systems result in a bimodal distributionof wind directions, mostly from south-west or from north-east.Based on studies relating weather conditions and migratory inten-sities (Erni et al., 2002; Zehnder et al., 2001a, 2001b), we estimatedthat within the study area 50% of the migrants selected weak windconditions (<5 m s�1), 30% moderate tail-winds (5–10 m s�1) and20% migrate under moderate head-winds (5–10 m/s). Becauseunfavourable headwinds predominate during migration acrossSwitzerland, there are always some movements into light head-winds. Due to low migration intensities, rare conditions like strongwinds (>10 m s�1) and winds from directions other than south-west or north-east were disregarded. For each of the three windscenarios, the birds’ flight directions were implemented as proba-bilities for flying into one of the neighbouring cells (detailed prob-abilities in the Supplementary material Table A2).

2.3. Effect of topography

Hills and mountains exceeding a bird’s flight altitude force iteither to gain height or to change direction. It is unknown at whatdistances flying birds are taking note of obstacles ahead and initi-ate an adequate response. In our model the effect of topographydepends on the range we expect a bird is taking into account to re-act to obstacles ahead. Our assumptions are based on single birdstracked in radar studies (Liechti and Bruderer, 1986). We used amoving window to calculate the altitude in the expected flightdirection, which might be relevant for the reaction of a bird. Thesize of the moving window represents the range a bird might con-sider. We tested different measures to represent the altitude with-in this range (e.g. mean, max, quartiles) and simulated variousranges for the moving window (1–400 km2). Based on the magni-tude of the difference between the bird’s flight altitude in the cellof origin and the height of the neighbouring landscape, respec-tively (range of the moving window), we applied a weight factor

Page 3: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

26 F. Liechti et al. / Biological Conservation 162 (2013) 24–32

to each of the eight directional probabilities (climb function). Theclimb function provides a coefficient between 0 and 1, followinga sigmoid curve (1 = climb, 0 = detour; for mathematical detailssee Supplementary material Fig. A3). We tested different valuesfor the parameters of the sigmoid curve (Supplementary materialTable A1). The elevation data were provided by the GTOPO30 dig-ital elevation model (USGS, 1996).

2.4. Flight altitude

Migrants generally climb up to altitudes where wind conditionsare more favourable than below (Mateos and Liechti, 2012). Thus,in headwinds they fly lower than under tailwind conditions. Withthe current model approach we could not take into account verticalwind profiles. Therefore, mean flight altitude was derived onlyfrom the fact that birds moving from a low grid cell to a higherone must climb. At the start, a parameter was set to assign all birdsto an initial flight altitude (Supplementary material Table A1).Afterwards, the mean flight altitude was calculated for each grid-cell based on the flight altitudes in the surrounding eight cellsweighted by the corresponding probabilities for moving into thefocal cell. To take into account that birds do not only climb to crossmountain areas, but also may descend behind mountain ranges(mainly in headwinds), we included a parameter allowing the birdsto descend by a given rate (fly-down rate). Otherwise, there wouldbe no effect of topography after they have crossed the highestmountain ranges. Within the simulation, birds were not allowedto descend below the initial height above ground given at take-off.

2.5. Model runs

Starting the simulation with 100 birds per grid-cell turned outto be sufficient to achieve stable and smooth probabilities. A sim-ulation consisted of a defined number of iterations. In every singleiteration process, the probability of occupancy and the mean flightaltitude were calculated for each grid cell. We ran various param-eter combinations to investigate the sensitivity of each parameter(see above) and to find a set which could reproduce results inagreement with the tuning data (see below). It turned out that

Table 1Migratory intensities in MTR [birds h�1 km�1] observed by fixed-beam radar at four differeFigs. 5 and 6). The calculations are based on means per day (from midnight to midnight).

Site coordinates (altitude) Observation period (N days withmeasurements)

Central Alps 46.6�N, 8.6�W (2140 masl)

4 August – 18 October 2010(51.1)9 March – 7 June 2010(82.9)

Jura 47.2�N, 7.0�W (1200 m asl) 11 August – 15 November 2010(94.8)23 February – 2 June 2011(87.0)

SW Alps 46.6�N, 8.6�W (1677 m asl) 6 August – 29 October 2007(82.1)

Plateau 47.75�N, 9.0�W (424 m asl) 12 August – 6 November 2008(84.6)

Prealps 47.2�N, 7.3�W (1200 m asl) 25 August – 23 October 2009(52.3)

a Height interval of wind turbines above ground level (agl).b 25 August to 18 October, 9 March to 2 June, resp.

700 iterations were sufficient to achieve a stable distribution ofprobabilities and mean flight altitudes. Each simulation producedtwo grid layers, the probabilities representing the migratory inten-sities and the mean flight altitudes. For the parameter representingthe height in the moving window, the 3rd quartile turned out togive the most plausible results. The different parameter settingsused for the analysis and the visual results of the sensitivity anal-ysis are given in the Supplementary material (Table A1). Obviously,the distribution of flight directions derived from the different windconditions had a major impact on the outcome (cf. Supplementarymaterial Fig. B8). Not surprisingly, the model was also highly sen-sitive to the function used to consider the effect of topography (cf.Supplementary material Figs. A4 and A7). The model was less sen-sitive to initial flight altitude, but obviously topographical featuresbehind the first high mountain ranges had an increasing impactwith increasing fly-down rate.

For the final results we ran the simulation three times with aspecific parameter set for each season (see results), using the dis-tribution of flight directions for weak winds, moderate tailwindsand moderate headwinds, respectively (see above).To provide aseasonal density map of bird migration, we combined the threesimulations by weighting them 5:3:2 (see above).

2.6. Model – tuning

Bruderer (1996) presented maps of relative autumn migrationintensities across Switzerland for three different wind conditions.These maps, drawn by hand, integrated the former expert knowl-edge based on long-term field observations and numerous radarstudies with airport surveillance radar as well as with short-rangetracking radar. A moon watching campaign in 1994/1995 providedsimultaneous snapshots of migration intensities for a few nightsacross the whole study area (Liechti et al., 1996a, 1996b). A quan-titative comparison with these kinds of published data was notpossible. Therefore, we qualitatively compared the relative geo-graphical intensity patterns from the sketches (Bruderer, 1996)and graphs (Liechti et al., 1996a, 1996b) with the intensity mapsfrom the simulations.

nt sites in Switzerland and one site nearby, in Germany, representing the Plateau (cf.

All heights <200 m agla

Mean (SD) max Percentage max

Completeperiod

Overlappingperiodb

Completeperiod

Overlappingperiodb

70 (93) 97 (97) 10% 10%427 427 43 4376 (59) 80 (58) 14% 14%204 204 47 47

281 (262) 381 (281) 43% 41%1166 1166 394 394315 (443) 316 (417) 26% 27%2188 1937 531 531

388 (350) 466 (382) 32% 29%1584 1584 959 398

510 (763) 605 (860) 20% 20%6251 6251 1832 1832

1148 (600) 1168 (603) 8% 8%2890 2890 236 227

Page 4: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

F. Liechti et al. / Biological Conservation 162 (2013) 24–32 27

Furthermore, we used seasonal averages from local radar datacollected recently at different sites in the study area(Komenda-Zehnder et al., 2010). We gathered radar data of fivemain migratory periods in autumn and in spring (Table 1). Theseradar observations took place at four sites in Switzerland andone close to the border, at Lake of Constance (Table 1, Fig. 1).During one autumn season, radar measurements were runningsimultaneously at two sites (Central Alps and Jura). The monitoringwas accomplished by radar systems specifically designed forquantifying bird migration (Schmaljohann et al., 2008; Zaugget al., 2008; Bruderer et al., 2012). For the comparison of intensitiesbetween these sites, we considered only the seasonal periods, inwhich observations overlapped at all sites. The model providesthe average migratory intensity over many years, whereas themeasurements were recorded in a single year only. Therefore, itwas not possible, and also not expected to adjust the model param-eters to achieve the same densities. However, we adjusted themodel parameters until we achieved the same relative sequencein migratory intensities for the observed and simulated intensities.

2.7. Software

The simulation was run with R 2.14.1 (R Development CoreTeam, 2011) and required the package ‘‘raster’’ (Hijmans and vanEtten, 2012). We visualized the generated density and altitudinallayers in ArcGIS (ArcGIS Desktop 10 Service Pack 2, 1999–2010ESRI Inc.).

3. Results

By varying only a few parameters, the model was able to pro-duce distinct patterns of migratory intensities, closely matchingthe main migratory flyways of the empirical findings (Fig. 2; cf.Supplementary material Figs. B1 and B2 from Bruderer, 1996 andLiechti et al., 1996a, 1996b). Weak winds entail high migratoryintensities throughout the Plateau, which was achieved by (1) alarge moving window, (2) a medium climbing rate and (3) a slight

A

C

100 km

100 km

1 2 3 4 5 6 7 8 9 10deciles

Fig. 2. Simulated intensities of broad front migration in autumn. The migratory intensitiintensities that result under the three wind situations predominantly selected by pasmoderate head wind) and D the intensities of the complete autumn migration period, cavolume observed under the assumed conditions. The boarder of Switzerland is marked wshading. The grey lines in A, B and C represent the interpretation of the migratory flow

fly-down rate after crossing mountain ranges (Fig. 2A). With tailwinds, bird migration concentrated along the northern border ofthe Alps and a large proportion entered the valleys and mountainranges of the Alps (Fig. 2B). This was achieved by (1) increasingthe directional concentration in combination with (2) a strongertendency to climb, (3) a smaller moving window and (4) a zerofly-down rate. Finally, head winds forced the birds to enter the al-pine valleys and produced high concentrations within a few valleys(Fig. 2C). This was achieved in the model by (1) increasing the scat-ter in flight directions, (2) a low tendency to climb, (3) a smallmoving window and (4) a strong-fly down rate. The complete set-ting of the parameters for these results is given in the Supplemen-tary material (Table A2). The same settings were selected for springmigration except that directions were opposite. For all three windconditions, the simulation resulted in a concentration of springmigration along the southern border of the Jura Mountains(Fig. 3). Under weak wind conditions and even more with tail-winds, migratory intensity was low over the main parts of the Alpsexcept for some channelling effects among the main valleys in thesouth eastern corner (Fig. 3A and B). With headwinds, springmigration was concentrated strongly within the valleys all acrossthe mountain range of the Alps (Fig. 3C).

The results of the three wind conditions were pooled within onemap, separately for autumn and for spring (Figs. 2D and 3D). Eachwind condition was weighted according to its estimated frequencyof occrurrence (see methods). In autumn, migratory intensity washighest along the northern border of the Alps, it was half as inten-sive in the Plateau, approximately one third in the Jura and SWAlps and approximately one tenth in the Central Alps.

4. Application

We used the radar observations to convert the relative intensi-ties of the simulation into absolute intensities. A common measurefor bird migration is the number of birds crossing a virtual line of1 km perpendicular to the migratory direction within 1 h(MTR = Migration Traffic Rate). The seasonal mean migratory

B

D

100 km

100 km

es are classified in deciles (where 10 is the most intense). The illustrations show theserines for autumn migration (A weak wind, i.e. <5 m/s, B moderate tail wind, Clculated by combining the result of the three situations according to the migratoryith a dark line, rivers and lakes are highlighted grey. The topography is depicted byby Bruderer for these wind conditions (1996, Supplementary material Fig. B1).

Page 5: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

A B

C D

100 km 100 km

100 km 100 km

1 2 3 4 5 6 7 8 9 10deciles

Fig. 3. Simulated intensities of broad front migration in spring (as in Fig. 2).

28 F. Liechti et al. / Biological Conservation 162 (2013) 24–32

intensities observed with radar ranged from 100 MTR to 1200 MTR.The relative densities derived from the simulations were linearlytransformed to this range (Fig. 4).

4.1. Collision risk estimation

The wind turbines most likely used in future wind farms inSwitzerland have an overall height of some 140 m to 160 m and ro-tor blades a length of about 40 to 50 m. The obstacle formed by thewhole structure of a wind turbine adds up to approx. 500 m2.Hence, if we imagine one wind turbine placed on a hypotheticalline of 1 km, this obstacle covers 0.25% of the area in the heightinterval of 0–200 m above ground level (agl). Observations of the

100 km

A

100 km100 km

C

Fig. 4. Absolute migratory intensities in MTR [birds h�1 km�1] after calibration of the sim(C) densities of both migratory periods combined by weighting the intensities with theobservations used for the calibrations are given in Table 1.

reaction of migrant birds approaching wind farms indicate that90% of the birds manage to avoid them (Desholm, 2006). Therefore,we roughly defined the collision rate for birds with one wind tur-bine to be 0.025% of the migration traffic rate within the 200 mheight interval swept by the turbine.

To estimate the effect of shutting down wind turbines whenmigration is very intense, we assumed that birds only collide witha rotating wind turbine, but not at all when the blades stand still.The calculations were based on the assumption that wind turbineswere only allowed to be operational only when migratory intensitywas below a given threshold. The threshold was set to 10, 20, 50,100, 500 and 1000 MTR, respectively. We used our data from radarobservations to calculate a potentially remaining collision rate at a

100 km

B

0 – 100 MTR

101 – 200 MTR201 – 300 MTR

301 – 400 MTR401 – 500 MTR

501 – 600 MTR601 – 700 MTR

701 – 800 MTR801 – 900 MTR

> 900 MTR

ulated data with radar measurements: (A) autumn migration, (B) spring migration,estimated migratory volumes by 2 (autumn) to 1 (spring). The results of the radar

Page 6: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

Fig. 5. Yearly number of potential bird collisions with a single wind turbine inrelation to a hypothetical threshold determining the proposed shut down regime.Examples from five sites with radar observations are presented (cf. Table 1). Thecalculations are based on the distribution of mean MTR < 200 m agl per 3 h (for nsee legend). With an increasing threshold the deactivation time of a turbinedecreases (see Fig. 6), which puts more and more birds at risk. In our example theaverage collision rate for the five sites would be 10 birds per year and turbine (blackdot with range bar), if wind turbines would be shut down at a threshold of MTR = 50[birds h�1 km�1].

Fig. 6. Proportion of deactivation time of a hypothetical wind turbine during oneyear in relation to the threshold level of migratory intensity (MTR = birds km�1 h�1)that is set to shut down the turbine to reduce the collision rate. The T-bar illustratesthe range of the proportion of deactivation time for the five sites if the thresholdlevel is set at MTR = 50. This would lead to reduction of operation time of 10% onaverage.

F. Liechti et al. / Biological Conservation 162 (2013) 24–32 29

hypothetical wind turbine for each of the six thresholds. For eachsite and season, we first calculated migratory intensities <200 magl in 3-h intervals and then calculated the remaining mean MTRof these 3-hourly values after the exclusion of all values abovethe selected threshold. Based on the remaining intensity, we calcu-lated for each site and threshold a seasonal (100 days) and finally ayearly number of potential collisions at a single hypotheticalturbine.

Based on the frequency of how often a 3-hourly mean MTR(<200 m agl) was above the given threshold we calculated the timea hypothetical wind turbine would remain closed down at this site.We assumed that the calculated proportions are representative forthe main migratory period, lasting approximately 200 days a year.However, only for two sites data were available for spring and au-tumn (Jura and Central Alps). In order to get the proportion ofdeactivation time per year, we multiplied the results by the factor200/365.

Based on official statements from Swiss authorities, we pre-sume that within the next two decades up to 1000 wind turbinesmight be installed on Swiss territory. Inevitably, this leads to a cer-tain number of additional casualties of birds by wind turbines. It isnot known whether or not this will be additional or compensatorymortality (Lebreton, 2005). At least for spring migration, we mustassume that most of the mortality will be additional, because themigrants have already survived the non-breeding period and areclose to reproduction. To apply a threshold at which wind turbinesshould be stopped, we first had to agree on a number of acceptablefatalities. The number of bird strikes at existing man-made con-structions is estimated to be huge (Kerlinger et al., 2010; Longcoreet al., 2013). For Switzerland we assume that more than a millionbirds collide with buildings, cars and other man-made construc-tions (Schmid et al., 2008). Based on this background, we assessedthat the acceptable number of additional collisions with windfarms should not exceed 1% of the existing fatalities, to avoid anyadditional effects that might be critical for bird populations. Thus,we set the threshold for a tolerable number of additional bird casu-alties due to wind turbines to be 10,000 per year. With the ex-pected 1000 wind turbines, we ended up with an acceptablecollision number of 10 birds per turbine per year.

4.2. Collision risk mitigation

The number of potential bird strikes at wind turbines washighly dependent on the site and increased with threshold levels(Fig. 5). However, the number of estimated collisions did not in-crease linearly but levelled off at all sites. Accepting a mean of10 bird strikes leads to a threshold MTR level of 50 (birds h�1 -km�1) within the 200 m height interval covered by the blades ofa turbine.

The proportion of deactivation time of turbines dependsstrongly on the site and decreased with increasing threshold level(Fig. 6). A threshold level set at 50 MTR would reduce the operationtime by 2% in the Central Alps, and by 31% at the northern border ofthe Alps. The loss of energy production would be substantiallysmaller than the loss in operation time, because bird migrationup to 200 m occurs mainly on days with rather weak winds(<5 m/s), when energy production by wind turbines is expectedto be low (Erni et al., 2002).

4.3. Collision risk map

The aim of the collision risk map was to visualize areas of highand low risks for migrating birds and to facilitate planning pro-cesses of wind farms by denominating areas of heavy conflicts.For this risk assessment based on our model, we calculated themigratory intensity within the height interval of wind turbines,

i.e. below 200 m agl. To derive the migration intensity below200 m agl, we assumed a simple uniform height distribution fromground level up to twice the height of the simulated mean flightaltitude, and then calculated the proportion of MTR within thisheight interval.

If we consider only the migratory intensities in the altitudinalrange of hypothetical wind turbines, the migratory intensities asmeasured by radar, were highest in the SW Alps and the Jura,25–33% less in the Plateau and SW alps and less than 10% in thein the Central Alps. Proportions were similar in spring, althoughonly data for the Central Alps and the Jura were available

Page 7: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

B

A

Fig. 7. Migratory intensities <200 m above ground level (A) illustrated as absolute migratory intensities (MTR [birds h�1 km�1]) and (B) classified in three levels of estimatedcollision risk of migrants with wind turbines (green: low risk, yellow: medium risk, orange: high risk).

30 F. Liechti et al. / Biological Conservation 162 (2013) 24–32

(Fig. 7A). In view of actions required we reclassified the data inthree levels. These define the actions required for future windfarms: (1) areas with low collision risks (<10 collision expectedper year and wind turbine), where no action is needed with respectto bird migration, (2) areas with medium collision risks (10–20),where some actions for mitigation are required, (3) areas with highcollision risks (>20), where significant measures are required, suchas switching off wind turbines during intense migration.

High collision rates were predicted in all pre-alpine regions, abroad band in northwestern Switzerland and for most mountaintops in the Jura and the Alps. Collision rate was predicted to beminimal in the lowest areas of the Plateau (green area) and inthe inner-alpine valleys (Fig. 7B). The transition zone between highand low collision risks is mostly narrow, except for the wide areawith intermediate collision risks across the eastern Plateau.

5. Discussion

5.1. modelling broad front migration

Broad front migration as treated in this study is typical for thevast majority of migratory birds. It does, however not include themovement patterns of soaring birds, which concentrate their fly-ways to specific local updrafts. To our knowledge, this is the firstattempt to present intensities of broad front migration over a largespatial and temporal scale using computer simulation. Up to now,overviews on the course of broad front migration had been the re-sults of compiling the knowledge of migration in an ‘‘expert knowl-edge sketch’’ (Toschi, 1939; Schüz, 1971; Lövei, 1989; Bruderer,1996; Bruderer and Liechti, 1999). Our model integrates theknowledge on the behaviour of migrant birds accumulated overmore than four decades (Baumgartner and Bruderer, 1985;Bruderer, 1996, 1997; Bruderer and Liechti, 1990, 1998). The mod-el resulted in differentiated maps of migratory intensities acrossthe very pronounced topography of Switzerland. With only slightchanges in the parameter settings, it was possible to achieve a highvariability in spatial intensity patterns and to reconstruct the mainfindings of the maps compiled by Bruderer (1996). Moreover, themodels’ results were in general agreement with field observations

collected by radar, infrared camera or moon watching techniquesall over Switzerland (Liechti et al., 1996a, 1996b; Zehnder et al.,2001a, 2001b). There were no studies available to tune the resultsfor spring simulations in the same way as for autumn.

However, we acknowledge that the model has its limitations andthat the accuracy of the modelled migration patterns, mainly withinthe inner-alpine area, might be critical. The setting of the parameterswas kept constant throughout the study area, ignoring completelythat birds might adapt their behaviour according to local featuresor as a result of the previous history of their flight. For example, itmight be relevant whether they are approaching the first mountainranges coming from the Plateau or have already passed the firstridges (Liechti and Bruderer, 1986). In the model, all birds start ona line. However, in nature birds start from many different roostingsites often leading to some concentrations, even at the beginningof their flight. To account to some extent for this simplification, wereduced the number of birds taking off in spring from the mountainranges of the Alps (south of Switzerland). Moreover, migrant birdsshow a remarkable ability to select flight altitudes with favourabletail winds (Williams and Williams, 1978; Bruderer et al., 1995;Liechti, 2006; Schmaljohann et al., 2007; Mateos and Liechti,2012). Once they find such a favourable wind layer, they may remainat this level regardless of the topography underneath. Winds wereintegrated in the model inducing a change in direction preferencesof the birds, but not as a criterion to select flight altitude. As a result,flight altitude only depends on topography. Therefore, height aboveground is generally low before a mountain range, but high behindthe first ridges. However, with the parameter fly-down rate we doaccount for the general wind situation with a higher rate underheadwind conditions. Thus, flight altitude decreases rapidly behindridges and passes (cf. Figs. 2C and 3C). These parameter settings arebased on qualitative results achieved from tracking single birds byradar within the Alps (Bruderer, 1982). Before we can provide a realquantitative validation of the model, we need to implement thespatial and temporal pattern of real wind conditions across the studyarea. Only then, can we compare predictions directly with radarmeasurements. Such a modelling approach would need to includealso vertical wind profiles, because flight altitude is strongly influ-enced by the level of favourable winds (Bruderer et al., 1995; Mateosand Liechti, 2012). This would allow a more site specific simulation

Page 8: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

F. Liechti et al. / Biological Conservation 162 (2013) 24–32 31

of flight altitude, but also of flight directions. At the time, we do nothave the possibility for such simulations on a large scale, becausethis requires computer capacities far beyond our possibilities. How-ever, a next step will be to validate the model quantitatively forsmall areas, by simulating temporal patterns at single sites with realwind data and comparing the outcome with local radarobservations.

5.2. Application

The sensitivity maps are planning tools to inform decision mak-ers at the very beginning of a wind farm project about potentialconflicts with migrating birds. The main consequence of the pro-posed shut-down algorithm would be that sites with frequentlyhigh migration intensities should not be considered as wind farmsites, because they are less economic than sites with similar windconditions, but lower migration intensities. The sensitivity maphas been made freely available and can be downloaded fromthe web (http://www.vogelwarte.ch/vogel-und-windkraftanlagen.html). However, at the time being, the map should be used as a toolto assess the potential risk for non-soaring migrants, and does notreplace any part of a site specific risk assessment.

Our aim was to model the mass movement of the mainly noc-turnal broad front migration and therefore, the results are notcomparable with other studies investigating the collision risk forindividual large birds foraging or migrating mainly by soaringflight (Bright et al., 2008; Douglas et al., 2012 and references there-in). Nevertheless, within mountainous areas concentrations ofdiurnal soaring migrants and broad front migration may occur atidentical locations (e.g. passes). Because there is hardly any knowl-edge on whether and how small diurnal and nocturnal migratorybirds avoid collisions, we decided to use a rough estimate for thecollision risk rather than sophisticated models, developed formovements of large birds like raptors or geese (e.g. Band et al.,2007). We neglected several effects on collision rates, like the ef-fect of bird size, flight speed or speed of the turbine blade, becausethese have a minor impact on collision rate. To improve estimatesof collision rates, we strongly suggest to study the collision fatali-ties in combination with height specific radar measurements,rather than implementing new models for collision rates of smallbirds. These kinds of studies are still lacking.

To implement a shut-down regime it is essential to define athreshold for bird migration intensity, above which wind turbinestrigger a shut down. To define a biologically meaningful threshold,we should know the effect of the expected additional mortalitycaused by wind turbines on the population demography of the spe-cies involved. However, for most bird species migrating throughSwitzerland these parameters are barely known. We cannot seethat sufficient information on this subject will be available withindue time. Therefore, we are convinced that there is no way out forus, as the experts in the field of bird migration, to take a pragmaticapproach with those data that are available now. All other partiesinvolved, like wind farm developers or government agencies, arenot in the position to propose a more reliable risk assessment.We therefore propose that an acceptable number of additionalfatalities by wind turbine should be about two orders of magni-tudes below casualties caused by tall man-made constructions.

To calculate collision risk based on migratory intensity, we esti-mated the proportion of birds which pass through the area of theturbine and may finally collide. Bird behaviour to avoid collisionshas a strong influence on the final mortality rates (Chamberlainet al., 2006). Currently, studies quantifying avoidance behaviourof passerines confronted with wind turbines are completely lack-ing. Recent investigations have mostly been performed in low alti-tude or topographically uniform areas (e.g. plains, onshore oroffshore), and focused on large bird species like waterfowl or rap-

tors, and reported avoidance rates >99% (Chamberlain et al., 2006).As many passerines migrate at night, avoiding obstacles might bemuch more difficult than in large diurnal birds, and we thereforehave chosen an arbitrary factor of 90%.

To reduce the mortality rate during times with high migratoryintensities, we propose to shut down the wind turbines. In reality,the structure of a wind turbine itself still represents a collision risk,but probably one much lower than that for a structure with rotat-ing blades. On average, a mortality rate of 10 birds per year andwind turbine should be tolerable. However, a local assessment isalways needed because it cannot be ruled out that, at a particularsite, 10 fatalities of a rare species might be a real threat for the spe-cific bird population. In Switzerland the authorities are planning toinclude the possibility of a shut-down regime in the list of mea-sures reducing casualties.

We are aware of the fact that our recommendation for a mitiga-tion application is based much more on expert assessment thanhard facts. But the wind energy industry is growing fast and weneed mitigation actions now to approve wind as real ‘‘green’’ en-ergy. The model presented is a tool which can be used and im-proved by anyone and may stimulate monitoring and large scalemodelling mass movements. Continent-wide bird migration mon-itoring is likely in the near future. As shown by Doktor et al. (2011)the existing European wide network of weather radars could pro-vide the spatio-temporal pattern of the large scale migratorymovements across Europe in almost real-time. This would allowto improve and expand our proposed methodology into a univer-sally accepted method of mapping the risk of wind turbines onmigratory birds.

Acknowledgments

This study was financed by the Swiss Federal Office for Environ-ment. We thank Silke Bauer, Valère Martin and Maria Mateos fortheir initial input in the development, programming and tuningof the model. We thank all the people involved in the radar obser-vation and data analysis, particularly Janine Aschwanden, DieterPeter and Herbert Stark. We acknowledge Thomas Steuri for thedevelopment of the sophisticated radar technique and Erich Bäch-ler who contributed specialized software. Lukas Jenni, GilbertoPasinelli and Luc Schifferli helped to improve the manuscript.Dan Chamberlain and two anonymous referees made valuablecomments on a former version of the manuscript.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.biocon.2013.03.018.

References

Alerstam, T., 1990. Bird Migration. Cambridge University Press, Cambridge.Band, W., Madders, M., Whitfield, D.P., 2007. Developing field and analytical

methods to assess avian collision risk at wind farms. In: De Lucas, M., Janss, G.,Ferrer, M. (Eds.), Birds and Wind Power. LynxEdicions, Barcelona, pp. 259–275.

Baumgartner, M., Bruderer, B., 1985. Radarbeobachtungen über die Richtungen desnächtlichen Vogelzuges am nördlichen Alpenrand. Ornithol. Beob. 82, 207–230.

Bright, J., Langston, R., Bullman, R., Gardner, S., Pearce-Higgins, J., 2008. Map of birdsensitivities to wind farms in Scotland: a tool toaid planning and conservation.Biol. Conserv. 141, 2342–2356.

Bruderer, B., 1982. Do migrating birds fly along straight lines? In: Papi, F. (Ed.),Avian Navigation. Springer-Verlag, Berlin, Heidelberg, pp. 3–14.

Bruderer, B., 1996. Vogelzugforschung im Bereich der Alpen 1980–1995. Ornithol.Beob. 93, 119–130.

Bruderer, B., 1997. The study of bird migration by radar. Part 2: Majorachievements. Naturwissenschaften 84, 45–54.

Bruderer, B., Jenni, L., 1990. Migration Across the Alps. In: E. Gwinner (Ed.). BirdMigration.

Page 9: Modelling the spatial concentrations of bird migration to assess conflicts with wind turbines

32 F. Liechti et al. / Biological Conservation 162 (2013) 24–32

Bruderer, B., Liechti, F., 1990. Richtungsverhalten nachtziehender Vögel inSüddeutschland und der Schweiz unter besonderer Berücksichtigung desWindeinflussses. Ornithol. Beob. 87, 271–293.

Bruderer, B., Liechti, F., 1998. Intensität, Höhe und Richtung von Tag- und Nachtzugim Herbst über Südwestdeutschland. Ornithol. Beob. 95, 113–128.

Bruderer, B., Liechti, F., 1999. Bird migration across the Mediterranean. In: Adams,N.J., Slotow, R.H. (Eds.), Proc. 22nd Int. Ornithol. Congr. Durban, 1998, pp. 1983–1999.

Bruderer, B., Underhill, L.G., Liechti, F., 1995. Altitude choice by night migrants in adesert area predicted by meteorological factors. Ibis 137, 44–55.

Bruderer, B., Steuri, T., Aschwanden, J., Liechti, F., 2012. Vom militärischenZielfolgeradar zum Vogelradar. Ornithol. Beob. 109, 157–176.

Chamberlain, D.E., Rehfisch, M.R., Fox, A.D., Desholm, M., Anthony, S.J., 2006. Theeffect of avoidance rates on bird mortality predictions made by wind turbinecollision risk models. Ibis 148, 198–202.

Desholm, M., 2006. Wind Farm Related Mortality Among Avian Migrants – ARemote Sensing Study and Model Analysis. Dissertation Univ., Copenhagen.

Doktor, A.M., Liechti, F., Stark, H., Delobbe, L., Tabary, P., Hollemann, I., 2011. Birdmigration flight altitudes studied by a network of operational weather radars. J.R. Soc. Interface 8, 30–43.

Douglas, D.J.T., Follestad, A., Langston, R.H.W., Pearce-Higgins, J.W., 2012. Modelledsensitivity of avian collision rate at wind turbines varies with number of hoursof flight activity input data. Ibis 154, 858–861. http://dx.doi.org/10.1111/j.1474-919X.2012.01239.x.

Drewitt, A.L., Langston, R.H.W., 2006. Assessing the impacts of wind farms on birds.Ibis 148, 29–42.

Erni, B., Liechti, F., Underhill, L., Bruderer, B., 2002. Wind and rain govern theintensity of nocturnal bird migration in central Europe – a log-linear regressionanalysis. Ardea 90, 155–166.

Global wind report, 2011. <http://www.gwec.net/fileadmin/documents/NewsDocuments/Annual_report_2011_lowres.pdf> (accessed 08.06.12).

Hijmans, Robert J., van Etten, Jacob, 2012. Raster: Geographic analysis and modelingwith raster data. R Packageversion 2.0-08. <http://CRAN.R-project.org/package=raster>.

Hüppop, O., Dierschke, J., Exo, K.M., Fredrich, E., Hill, R., 2006. Bird migration studiesand potential collision risk with offshore wind turbines. Ibis 148, 90–109.

Kerlinger, P., 1989. Flight Strategies of Migrating Hawks. University of ChicagoPress.

Kerlinger, P., Gehring, J.L., Erickson, W.P., Curry, R., Jain, A., Guarnaccia, J., 2010.Night migrant fatalities and obstruction lighting at wind turbines in NorthAmerica. Wilson J. Ornithol. 122, 744–754.

Komenda-Zehnder, S., Liechti, F., Jenni, L., 2010. Do bird captures reflect migrationintensity? – Trapping numbers on an alpine pass compared with radar counts. J.Avian Biol. 41, 434–444.

Lebreton, J.-D., 2005. Dynamical and statistical models for exploited populations.Aust. NZ J. Stat. 47, 49–63.

Liechti, F., 1993. Nächtlicher vogelzug im herbst über süddeutschland: winddriftund kompensation. J. Ornithol. 134, 373–404.

Liechti, F., 2006. Birds: Blowin’ by the wind. J. Ornithol. 147, 102–111.

Liechti, F., Bruderer, B., 1986. Einfluss der lokalen Topographie auf nächtlichziehende Vögel nach Radarstudien am Alpenrand. Ornithol. Beob. 83, 35–66.

Liechti, F., Peter, D., Lardelli, R., Bruderer, B., 1996a. Die Alpen, ein Hindernis imnächtlichen Breitfrontzug – eine grossräumige Übersicht nachMondbeobachtungen. J. Ornithol. 137, 337–356.

Liechti, F., Peter, D., Lardelli, R., Bruderer, B., 1996b. Herbstlicher Vogelzug imAlpenraum nach Mondbeobachtungen – Topographie und Wind beeinflussenden Zugverlauf. Ornithol. Beob. 93, 131–152.

Longcore, T. et al., 2013. Avian mortality at communication towers in the UnitedStates and Canada: which species, how many, and where? Biol. Conserv. 158,410.

Lövei, G.L., 1989. Passerine migration between the Palaearctic and Africa. Curr.Ornithol. 6, 143–174.

Mateos, M., Liechti, F., 2012. How do diurnal long-distance migrants select flightaltitude in relation to wind? Behav. Ecol. 23, 403–409.

Newton, I., 2008. The Migration Ecology of Birds. Academic Press, London.Packard, N.H., Wolfram, S., 1985. Two-dimensional cellular automata. J. Stat. Phys.

38, 901–946.Pennycuick, C.J., 1975. Mechanics of flight. In: Farner, D.S. (Ed.), Avian Biology.

Academic Press, New York, pp. 1–75.R Development Core Team, 2011. R: A Language and Environment for Statistical

Computing. R Foundation for Statistical Computing, Vienna.Austria. ISBN 3-900051-07-0, <http://www.R-project.org>.

Schmaljohann, H., Liechti, F., Bruderer, B., 2007. Songbird migration across theSahara: the non-stop hypothesis rejected! Proc. R. Soc. Lond. B 274, 735–739.

Schmaljohann, H., Liechti, F., Bächler, E., Steuri, T., Bruderer, B., 2008. Quantificationof bird migration by radar – a detection probability problem. Ibis 150, 342–350.

Schmid, H., Waldburger, P., Heynen, D., 2008. Vogelfreundliches Bauen mit Glas undLicht. Schweizerische Vogelwarte, Sempach.

Schüz, E., 1971. Grundriss der Vogelzugkunde. Verlag Paul Parey, Berlin undHamburg.

Toschi, A., 1939. La migrazione Degli Uccelli. Editrice Compositori, Bologna.USGS, 1996. GTOPO30. URL Earth Resources Observation and Science (EROS) Center,

<http://eros.usgs.gov/#/Find_Data/Products_and_Data_Available/gtopo30_info> (accessed 08.06.12).

Van Belle, J., Shamoun-Baranes, J., Van Loon, E., Bouten, W., 2007. An operationalmodel predicting autumn bird migration intensities for flight safety. J. Appl.Ecol. 44, 864–874.

Williams, T.C., Williams, J.M., 1978. An oceanic mass migration of land birds. Sci.Am. 239, 138–145.

Winkler, R., 1999. Avifauna der Schweiz. Ornithol. Beob., Beiheft 10, 252.Zaugg, S., Saporta, G., van Loon, E., Schmaljohann, H., Liechti, F., 2008. Automatic

identification of bird targets with radar via patterns produced by wing flapping.J. R. Soc. Interface 5, 1041–1053.

Zehnder, S., Åkesson, S., Liechti, F., Bruderer, B., 2001a. Nocturnal autumn birdmigration at Falsterbo, South Sweden. J. Avian Biol. 32, 239–248.

Zehnder, S., Liechti, F., Trösch, B., Bruderer, B., 2001b. Gibt es topographie-unabhängigen Nachtzug über den Alpen? Ornithol. Beob. 98, 215–222.


Recommended